Operation result error when run on pytorch:1.12.0-cuda11.3

[Liuyunming1999]

Thank you for your work！The results of my run on the dataset are shown in this figure. Obviously, I have encountered a problem. I hope you can help me with the answer. Thank you very much!

[mmmmimic]

Thank you for your question! Would you mind sharing the training log? I plan to try to debug the segmentation code after my vacation (April 10th), sorry for the bugs!

[Liuyunming1999]

log.txt
Here is my log file, thank you very much

[mmmmimic]

Hi! Sorry for the late reply. I have through the pipeline (data preparation + model training) again. Again, I failed to reproduce the error you encountered. I would suspect this is due to the environment. Therefore, I have updated my environment in the repository ('environment.yaml'). You may install it with conda env create -f environment.yaml. Please let me know if the model is not working under this environment - then it would be great if you can leave your email address for me to share the data I just generated today.

[suetme]

Hi, thank you for this excellent work.！When I trained the Partseg model using my own dataset, I encountered a situation where the acc was very low and the iou was very high. This is my training log：
Namespace(batch_size=24, dataset='MaizeSeg', dropout=0.5, epochs=2000, eval=False, exp_name='MaizeSeg_exp1', lr=0.001, model_path='', momentum=0.9, num_noise=100, num_points=4096, radius=0.01, seed=42, test_batch_size=24)
Using CPU
Namespace(batch_size=24, dataset='MaizeSeg', dropout=0.5, epochs=2000, eval=False, exp_name='MaizeSeg_exp1', lr=0.001, model_path='', momentum=0.9, num_noise=100, num_points=4096, radius=0.01, seed=42, test_batch_size=24)
Using GPU : 0 from 1 devices
Namespace(batch_size=24, dataset='MaizeSeg', dropout=0.5, epochs=2000, eval=False, exp_name='MaizeSeg_exp1', lr=0.001, model_path='', momentum=0.9, num_noise=100, num_points=4096, radius=0.01, seed=42, test_batch_size=24)
Using GPU : 0 from 1 devices
Namespace(batch_size=24, dataset='MaizeSeg', dropout=0.5, epochs=2000, eval=False, exp_name='MaizeSeg_exp1', lr=0.001, model_path='', momentum=0.9, num_noise=100, num_points=4096, radius=0.01, seed=42, test_batch_size=24)
Using GPU : 0 from 1 devices
Namespace(batch_size=16, dataset='MaizeSeg', dropout=0.5, epochs=2000, eval=False, exp_name='MaizeSeg_exp1', lr=0.001, model_path='', momentum=0.9, num_noise=100, num_points=4096, radius=0.01, seed=42, test_batch_size=24)
Using GPU : 0 from 1 devices
Namespace(batch_size=8, dataset='MaizeSeg', dropout=0.5, epochs=2000, eval=False, exp_name='MaizeSeg_exp1', lr=0.001, model_path='', momentum=0.9, num_noise=100, num_points=4096, radius=0.01, seed=42, test_batch_size=24)
Using GPU : 0 from 1 devices
Train 0, loss: 3.598473, train acc: 0.094918, train avg acc: 0.025307, train iou: 0.356386
Namespace(batch_size=16, dataset='MaizeSeg', dropout=0.5, epochs=2000, eval=False, exp_name='MaizeSeg_exp1', lr=0.001, model_path='', momentum=0.9, num_noise=100, num_points=4096, radius=0.01, seed=42, test_batch_size=16)
Using GPU : 0 from 1 devices
Namespace(batch_size=8, dataset='MaizeSeg', dropout=0.5, epochs=2000, eval=False, exp_name='MaizeSeg_exp1', lr=0.001, model_path='', momentum=0.9, num_noise=100, num_points=4096, radius=0.01, seed=42, test_batch_size=8)
Using GPU : 0 from 1 devices
Train 0, loss: 3.610561, train acc: 0.103682, train avg acc: 0.029265, train iou: 0.335014
Test 0, loss: 4.191408, test acc: 0.089150, test avg acc: 0.023810, test iou: 0.894737
Train 1, loss: 3.342831, train acc: 0.156049, train avg acc: 0.037794, train iou: 0.814118
Test 1, loss: 3.508400, test acc: 0.110152, test avg acc: 0.027457, test iou: 0.842105
Train 2, loss: 3.247782, train acc: 0.163670, train avg acc: 0.041938, train iou: 0.756361
Test 2, loss: 3.200593, test acc: 0.225303, test avg acc: 0.078232, test iou: 0.810978
Train 3, loss: 3.159302, train acc: 0.189925, train avg acc: 0.065864, train iou: 0.604232
Test 3, loss: 3.089429, test acc: 0.242104, test avg acc: 0.093007, test iou: 0.882348
Train 4, loss: 3.108613, train acc: 0.213755, train avg acc: 0.071538, train iou: 0.703454
Test 4, loss: 3.150975, test acc: 0.189556, test avg acc: 0.070814, test iou: 0.860397
Train 5, loss: 3.161339, train acc: 0.198765, train avg acc: 0.067291, train iou: 0.694777
Test 5, loss: 3.128125, test acc: 0.196854, test avg acc: 0.068024, test iou: 0.854817
Train 6, loss: 3.012451, train acc: 0.229540, train avg acc: 0.088954, train iou: 0.659943
Test 6, loss: 3.025875, test acc: 0.265946, test avg acc: 0.106141, test iou: 0.803291
Train 7, loss: 2.923977, train acc: 0.268329, train avg acc: 0.087133, train iou: 0.700083
Test 7, loss: 2.971429, test acc: 0.250520, test avg acc: 0.083275, test iou: 0.871449
Train 8, loss: 2.917160, train acc: 0.264699, train avg acc: 0.101344, train iou: 0.725877
Test 8, loss: 2.965597, test acc: 0.256425, test avg acc: 0.075944, test iou: 0.861432
Train 9, loss: 2.948400, train acc: 0.247255, train avg acc: 0.095647, train iou: 0.730285
Test 9, loss: 2.848340, test acc: 0.296464, test avg acc: 0.093425, test iou: 0.867808
Train 10, loss: 2.737956, train acc: 0.314266, train avg acc: 0.125697, train iou: 0.820346
Test 10, loss: 2.880172, test acc: 0.304161, test avg acc: 0.096272, test iou: 0.867830
Train 11, loss: 2.729066, train acc: 0.310389, train avg acc: 0.123528, train iou: 0.710061
Test 11, loss: 2.810601, test acc: 0.289615, test avg acc: 0.094828, test iou: 0.878062
Train 12, loss: 2.714566, train acc: 0.303599, train avg acc: 0.112171, train iou: 0.743845
Test 12, loss: 2.736460, test acc: 0.323981, test avg acc: 0.113096, test iou: 0.840451
Train 13, loss: 2.637697, train acc: 0.338244, train avg acc: 0.140140, train iou: 0.692961
Test 13, loss: 2.697827, test acc: 0.333194, test avg acc: 0.115578, test iou: 0.933891
Train 14, loss: 2.602361, train acc: 0.343745, train avg acc: 0.142981, train iou: 0.791191
Test 14, loss: 2.666919, test acc: 0.345151, test avg acc: 0.129192, test iou: 0.904366
Train 15, loss: 2.592404, train acc: 0.345591, train avg acc: 0.148484, train iou: 0.718633
Test 15, loss: 2.563002, test acc: 0.359870, test avg acc: 0.143987, test iou: 0.875381
Train 16, loss: 2.529455, train acc: 0.365177, train avg acc: 0.157687, train iou: 0.813444
Test 16, loss: 2.554642, test acc: 0.384740, test avg acc: 0.161831, test iou: 0.879140
Train 17, loss: 2.561692, train acc: 0.352779, train avg acc: 0.154625, train iou: 0.713150
Test 17, loss: 2.551154, test acc: 0.364412, test avg acc: 0.141967, test iou: 0.879140
Train 18, loss: 2.525412, train acc: 0.358914, train avg acc: 0.169307, train iou: 0.762349
Test 18, loss: 2.468505, test acc: 0.386571, test avg acc: 0.165154, test iou: 0.886651
Train 19, loss: 2.501144, train acc: 0.368867, train avg acc: 0.168237, train iou: 0.784222
Test 19, loss: 2.524918, test acc: 0.363782, test avg acc: 0.164509, test iou: 0.877528
Train 20, loss: 2.502066, train acc: 0.363811, train avg acc: 0.166691, train iou: 0.804927
Test 20, loss: 2.482035, test acc: 0.375206, test avg acc: 0.165309, test iou: 0.886651
Train 21, loss: 2.460626, train acc: 0.373458, train avg acc: 0.168725, train iou: 0.848704
Test 21, loss: 2.443182, test acc: 0.372057, test avg acc: 0.164562, test iou: 0.933891
Train 22, loss: 2.432817, train acc: 0.376337, train avg acc: 0.174111, train iou: 0.863090
Test 22, loss: 2.444168, test acc: 0.377943, test avg acc: 0.167577, test iou: 0.885313
Train 23, loss: 2.427435, train acc: 0.375429, train avg acc: 0.172887, train iou: 0.853903
Test 23, loss: 2.439689, test acc: 0.383712, test avg acc: 0.169169, test iou: 0.886651
Train 24, loss: 2.429235, train acc: 0.374706, train avg acc: 0.173577, train iou: 0.859281
Test 24, loss: 2.441073, test acc: 0.375051, test avg acc: 0.170098, test iou: 0.886651
Train 25, loss: 2.429076, train acc: 0.368645, train avg acc: 0.175210, train iou: 0.843714
Test 25, loss: 2.416055, test acc: 0.390143, test avg acc: 0.175161, test iou: 0.886651
Train 26, loss: 2.398142, train acc: 0.382281, train avg acc: 0.177022, train iou: 0.873366
Test 26, loss: 2.403808, test acc: 0.375128, test avg acc: 0.168920, test iou: 0.829978
Train 27, loss: 2.398264, train acc: 0.379312, train avg acc: 0.177938, train iou: 0.868480
Test 27, loss: 2.385745, test acc: 0.384175, test avg acc: 0.170974, test iou: 0.886651
Train 28, loss: 2.406543, train acc: 0.381017, train avg acc: 0.176022, train iou: 0.859754
Test 28, loss: 2.397959, test acc: 0.375970, test avg acc: 0.168955, test iou: 0.886651
Train 29, loss: 2.404365, train acc: 0.379997, train avg acc: 0.181768, train iou: 0.923678
Test 29, loss: 2.399358, test acc: 0.379799, test avg acc: 0.169440, test iou: 0.904366
Train 30, loss: 2.405236, train acc: 0.376435, train avg acc: 0.171544, train iou: 0.806487
Test 30, loss: 2.409859, test acc: 0.380962, test avg acc: 0.166969, test iou: 0.829975
Train 31, loss: 2.394001, train acc: 0.379034, train avg acc: 0.178828, train iou: 0.884045
Test 31, loss: 2.396502, test acc: 0.381097, test avg acc: 0.164969, test iou: 0.904366
Train 32, loss: 2.405356, train acc: 0.379988, train avg acc: 0.179629, train iou: 0.878071
Test 32, loss: 2.399034, test acc: 0.379401, test avg acc: 0.166680, test iou: 0.933891
Train 33, loss: 2.395335, train acc: 0.380540, train avg acc: 0.176929, train iou: 0.862521
Test 33, loss: 2.394332, test acc: 0.382947, test avg acc: 0.171452, test iou: 0.904366
Train 34, loss: 2.397363, train acc: 0.380476, train avg acc: 0.176916, train iou: 0.870394
Test 34, loss: 2.391332, test acc: 0.383063, test avg acc: 0.169512, test iou: 0.886651
Train 35, loss: 2.393285, train acc: 0.382959, train avg acc: 0.178195, train iou: 0.856768
Test 35, loss: 2.390840, test acc: 0.379459, test avg acc: 0.171195, test iou: 0.888162
Train 36, loss: 2.406141, train acc: 0.379810, train avg acc: 0.181469, train iou: 0.879207
Test 36, loss: 2.376310, test acc: 0.382138, test avg acc: 0.168966, test iou: 0.886651
Train 37, loss: 2.381790, train acc: 0.379559, train avg acc: 0.175265, train iou: 0.881616
Test 37, loss: 2.382012, test acc: 0.389411, test avg acc: 0.173328, test iou: 0.904366
Train 38, loss: 2.383590, train acc: 0.384686, train avg acc: 0.182388, train iou: 0.905409
Test 38, loss: 2.385729, test acc: 0.381007, test avg acc: 0.165375, test iou: 0.889699
Train 39, loss: 2.375693, train acc: 0.388804, train avg acc: 0.185673, train iou: 0.911564
Test 39, loss: 2.380262, test acc: 0.384778, test avg acc: 0.171467, test iou: 0.933891
Train 40, loss: 2.416964, train acc: 0.372155, train avg acc: 0.173330, train iou: 0.798207
Test 40, loss: 2.378633, test acc: 0.383718, test avg acc: 0.170555, test iou: 0.886651
Train 41, loss: 2.379409, train acc: 0.384820, train avg acc: 0.182597, train iou: 0.888276
Test 41, loss: 2.384936, test acc: 0.373285, test avg acc: 0.161690, test iou: 0.886651
Train 42, loss: 2.376884, train acc: 0.384344, train avg acc: 0.179950, train iou: 0.817345
Test 42, loss: 2.378827, test acc: 0.373053, test avg acc: 0.166366, test iou: 0.904366
Train 43, loss: 2.372270, train acc: 0.387563, train avg acc: 0.184846, train iou: 0.941538
Test 43, loss: 2.376033, test acc: 0.383339, test avg acc: 0.167375, test iou: 0.887665
Train 44, loss: 2.372946, train acc: 0.386139, train avg acc: 0.182460, train iou: 0.840083
Test 44, loss: 2.375479, test acc: 0.369006, test avg acc: 0.159184, test iou: 0.887617
Train 45, loss: 2.368091, train acc: 0.380900, train avg acc: 0.182783, train iou: 0.873458
Test 45, loss: 2.373236, test acc: 0.378116, test avg acc: 0.171416, test iou: 0.888008
Train 46, loss: 2.359972, train acc: 0.382694, train avg acc: 0.175575, train iou: 0.872781
Test 46, loss: 2.372729, test acc: 0.371074, test avg acc: 0.165064, test iou: 0.887617
Train 47, loss: 2.361814, train acc: 0.388320, train avg acc: 0.183813, train iou: 0.839497
Test 47, loss: 2.371212, test acc: 0.383282, test avg acc: 0.169889, test iou: 0.888548
Train 48, loss: 2.364406, train acc: 0.387898, train avg acc: 0.183284, train iou: 0.881579
Test 48, loss: 2.376586, test acc: 0.371704, test avg acc: 0.167472, test iou: 0.904366
Train 49, loss: 2.361022, train acc: 0.387168, train avg acc: 0.183698, train iou: 0.879221
Test 49, loss: 2.373726, test acc: 0.378206, test avg acc: 0.167010, test iou: 0.888277
Train 50, loss: 2.391375, train acc: 0.387218, train avg acc: 0.186030, train iou: 0.904826
Test 50, loss: 2.383692, test acc: 0.385177, test avg acc: 0.169463, test iou: 0.886651
Train 51, loss: 2.363602, train acc: 0.383139, train avg acc: 0.177373, train iou: 0.886680
Test 51, loss: 2.379164, test acc: 0.371730, test avg acc: 0.164626, test iou: 0.829076
Train 52, loss: 2.365342, train acc: 0.386497, train avg acc: 0.186809, train iou: 0.884005
Test 52, loss: 2.375016, test acc: 0.378174, test avg acc: 0.163704, test iou: 0.904366
Train 53, loss: 2.361375, train acc: 0.387506, train avg acc: 0.183708, train iou: 0.887293
Test 53, loss: 2.383640, test acc: 0.365691, test avg acc: 0.166198, test iou: 0.886454
Train 54, loss: 2.356995, train acc: 0.388261, train avg acc: 0.188733, train iou: 0.881652
Test 54, loss: 2.377177, test acc: 0.374794, test avg acc: 0.165637, test iou: 0.933891
Train 55, loss: 2.381183, train acc: 0.381945, train avg acc: 0.178373, train iou: 0.906061
Test 55, loss: 2.371465, test acc: 0.376041, test avg acc: 0.167376, test iou: 0.904366
Train 56, loss: 2.359288, train acc: 0.386281, train avg acc: 0.181335, train iou: 0.885356
Test 56, loss: 2.380269, test acc: 0.364033, test avg acc: 0.157207, test iou: 0.904366
Train 57, loss: 2.358563, train acc: 0.385697, train avg acc: 0.178793, train iou: 0.884809
Test 57, loss: 2.377831, test acc: 0.369147, test avg acc: 0.164310, test iou: 0.887617
Train 58, loss: 2.363497, train acc: 0.384930, train avg acc: 0.184073, train iou: 0.837592
Test 58, loss: 2.381068, test acc: 0.375116, test avg acc: 0.164612, test iou: 0.933891
Train 59, loss: 2.364529, train acc: 0.386634, train avg acc: 0.179892, train iou: 0.886610
Test 59, loss: 2.382034, test acc: 0.368023, test avg acc: 0.162109, test iou: 0.877985
Train 60, loss: 2.353246, train acc: 0.388895, train avg acc: 0.184278, train iou: 0.872951
Test 60, loss: 2.372546, test acc: 0.365485, test avg acc: 0.166944, test iou: 0.886454
Train 61, loss: 2.361241, train acc: 0.390353, train avg acc: 0.189083, train iou: 0.909834
Test 61, loss: 2.378233, test acc: 0.378990, test avg acc: 0.167776, test iou: 0.933891
Train 62, loss: 2.357041, train acc: 0.388537, train avg acc: 0.182928, train iou: 0.941941
Test 62, loss: 2.375372, test acc: 0.368003, test avg acc: 0.168236, test iou: 0.887617
Train 63, loss: 2.366135, train acc: 0.386297, train avg acc: 0.184446, train iou: 0.890337
Test 63, loss: 2.371402, test acc: 0.374692, test avg acc: 0.161655, test iou: 0.888775

[suetme]

Over the next iterations, the loss remains around 2.37, the acc remains around 0.38, the avgacc remains around 0.17, and the iou fluctuates between 0.85 and 0.93. By the way, my dataset only has 164 samples.
Test 1945, loss: 2.370986, test acc: 0.384566, test avg acc: 0.168906, test iou: 0.879140
Train 1946, loss: 2.317072, train acc: 0.397260, train avg acc: 0.195221, train iou: 0.889861
Test 1946, loss: 2.371344, test acc: 0.383127, test avg acc: 0.166341, test iou: 0.889699
Train 1947, loss: 2.315615, train acc: 0.396222, train avg acc: 0.192418, train iou: 0.889738
Test 1947, loss: 2.370504, test acc: 0.383436, test avg acc: 0.166659, test iou: 0.889699
Train 1948, loss: 2.315509, train acc: 0.397189, train avg acc: 0.193725, train iou: 0.889569
Test 1948, loss: 2.370401, test acc: 0.384566, test avg acc: 0.168906, test iou: 0.829975
Train 1949, loss: 2.316411, train acc: 0.396824, train avg acc: 0.192293, train iou: 0.885631
Test 1949, loss: 2.371268, test acc: 0.382980, test avg acc: 0.166189, test iou: 0.889699
Train 1950, loss: 2.320550, train acc: 0.396222, train avg acc: 0.193749, train iou: 0.912101
Test 1950, loss: 2.370480, test acc: 0.384566, test avg acc: 0.168906, test iou: 0.904366
Train 1951, loss: 2.320201, train acc: 0.396717, train avg acc: 0.196277, train iou: 0.842722
Test 1951, loss: 2.371929, test acc: 0.382620, test avg acc: 0.165818, test iou: 0.886651
Train 1952, loss: 2.318060, train acc: 0.396544, train avg acc: 0.193250, train iou: 0.942122
Test 1952, loss: 2.370918, test acc: 0.375765, test avg acc: 0.166597, test iou: 0.888351
Train 1953, loss: 2.318564, train acc: 0.396792, train avg acc: 0.193574, train iou: 0.889375
Test 1953, loss: 2.371295, test acc: 0.382755, test avg acc: 0.166494, test iou: 0.886651
Train 1954, loss: 2.317349, train acc: 0.397518, train avg acc: 0.193482, train iou: 0.889754
Test 1954, loss: 2.369916, test acc: 0.383217, test avg acc: 0.166434, test iou: 0.889699
Train 1955, loss: 2.318933, train acc: 0.395877, train avg acc: 0.191355, train iou: 0.873974
Test 1955, loss: 2.372064, test acc: 0.384566, test avg acc: 0.168906, test iou: 0.886651
Train 1956, loss: 2.316915, train acc: 0.396852, train avg acc: 0.192248, train iou: 0.842696
Test 1956, loss: 2.369206, test acc: 0.383391, test avg acc: 0.166613, test iou: 0.886651
Train 1957, loss: 2.320047, train acc: 0.396649, train avg acc: 0.193815, train iou: 0.889426
Test 1957, loss: 2.371206, test acc: 0.381964, test avg acc: 0.165144, test iou: 0.889699
Train 1958, loss: 2.318520, train acc: 0.396069, train avg acc: 0.194339, train iou: 0.942174
Test 1958, loss: 2.371605, test acc: 0.384566, test avg acc: 0.168906, test iou: 0.904366
Train 1959, loss: 2.317251, train acc: 0.397137, train avg acc: 0.193927, train iou: 0.911792
Test 1959, loss: 2.371200, test acc: 0.384566, test avg acc: 0.168906, test iou: 0.886651
Train 1960, loss: 2.321551, train acc: 0.395469, train avg acc: 0.195617, train iou: 0.906708
Test 1960, loss: 2.369627, test acc: 0.372616, test avg acc: 0.164244, test iou: 0.828898
Train 1961, loss: 2.319752, train acc: 0.397180, train avg acc: 0.192629, train iou: 0.889766
Test 1961, loss: 2.369780, test acc: 0.375212, test avg acc: 0.166027, test iou: 0.904366
Train 1962, loss: 2.316466, train acc: 0.397552, train avg acc: 0.192639, train iou: 0.911485
Test 1962, loss: 2.371272, test acc: 0.375662, test avg acc: 0.164019, test iou: 0.829310
Train 1963, loss: 2.317489, train acc: 0.396350, train avg acc: 0.191684, train iou: 0.889883
Test 1963, loss: 2.371849, test acc: 0.382639, test avg acc: 0.165838, test iou: 0.933891
Train 1964, loss: 2.321414, train acc: 0.396528, train avg acc: 0.195727, train iou: 0.911882
Test 1964, loss: 2.371795, test acc: 0.385273, test avg acc: 0.168554, test iou: 0.886651
Train 1965, loss: 2.323450, train acc: 0.397828, train avg acc: 0.193304, train iou: 0.911355
Test 1965, loss: 2.372183, test acc: 0.383551, test avg acc: 0.166778, test iou: 0.904366
Train 1966, loss: 2.318480, train acc: 0.395788, train avg acc: 0.191425, train iou: 0.942607
Test 1966, loss: 2.371025, test acc: 0.382729, test avg acc: 0.165930, test iou: 0.933891
Train 1967, loss: 2.317258, train acc: 0.395061, train avg acc: 0.190623, train iou: 0.889453
Test 1967, loss: 2.371001, test acc: 0.384464, test avg acc: 0.167719, test iou: 0.829975
Train 1968, loss: 2.318910, train acc: 0.395464, train avg acc: 0.192160, train iou: 0.889790
Test 1968, loss: 2.371105, test acc: 0.383583, test avg acc: 0.166811, test iou: 0.904366
Train 1969, loss: 2.319059, train acc: 0.396430, train avg acc: 0.193149, train iou: 0.875410
Test 1969, loss: 2.371892, test acc: 0.383384, test avg acc: 0.166606, test iou: 0.879140
Train 1970, loss: 2.321073, train acc: 0.397089, train avg acc: 0.195461, train iou: 0.911973
Test 1970, loss: 2.371132, test acc: 0.384566, test avg acc: 0.168906, test iou: 0.889699
Train 1971, loss: 2.316530, train acc: 0.397068, train avg acc: 0.193086, train iou: 0.890034
Test 1971, loss: 2.372177, test acc: 0.384566, test avg acc: 0.168906, test iou: 0.904366
Train 1972, loss: 2.319103, train acc: 0.396302, train avg acc: 0.191644, train iou: 0.817738
Test 1972, loss: 2.372295, test acc: 0.382973, test avg acc: 0.166182, test iou: 0.886651
Train 1973, loss: 2.317512, train acc: 0.397119, train avg acc: 0.191563, train iou: 0.912044
Test 1973, loss: 2.373002, test acc: 0.384412, test avg acc: 0.168706, test iou: 0.829975
Train 1974, loss: 2.319139, train acc: 0.394892, train avg acc: 0.192304, train iou: 0.889326
Test 1974, loss: 2.372499, test acc: 0.379144, test avg acc: 0.164934, test iou: 0.878635
Train 1975, loss: 2.319003, train acc: 0.396311, train avg acc: 0.193295, train iou: 0.885752
Test 1975, loss: 2.371416, test acc: 0.383513, test avg acc: 0.166739, test iou: 0.886651
Train 1976, loss: 2.318240, train acc: 0.396523, train avg acc: 0.194344, train iou: 0.911730
Test 1976, loss: 2.372317, test acc: 0.382851, test avg acc: 0.166056, test iou: 0.886651
Train 1977, loss: 2.322701, train acc: 0.395271, train avg acc: 0.193681, train iou: 0.839193
Test 1977, loss: 2.370505, test acc: 0.383795, test avg acc: 0.167030, test iou: 0.829975
Train 1978, loss: 2.316416, train acc: 0.396902, train avg acc: 0.192933, train iou: 0.885392
Test 1978, loss: 2.370879, test acc: 0.379035, test avg acc: 0.165498, test iou: 0.904366
Train 1979, loss: 2.316395, train acc: 0.396265, train avg acc: 0.192222, train iou: 0.889455
Test 1979, loss: 2.370659, test acc: 0.375032, test avg acc: 0.165842, test iou: 0.877985
Train 1980, loss: 2.316475, train acc: 0.396505, train avg acc: 0.193750, train iou: 0.885213
Test 1980, loss: 2.371453, test acc: 0.383699, test avg acc: 0.166931, test iou: 0.933891
Train 1981, loss: 2.315943, train acc: 0.396808, train avg acc: 0.191838, train iou: 0.889390
Test 1981, loss: 2.371143, test acc: 0.377332, test avg acc: 0.168214, test iou: 0.887617
Train 1982, loss: 2.316804, train acc: 0.396998, train avg acc: 0.193861, train iou: 0.942274
Test 1982, loss: 2.371442, test acc: 0.379594, test avg acc: 0.165903, test iou: 0.829517
Train 1983, loss: 2.317900, train acc: 0.396674, train avg acc: 0.193601, train iou: 0.889659
Test 1983, loss: 2.371251, test acc: 0.383282, test avg acc: 0.166500, test iou: 0.886651
Train 1984, loss: 2.315678, train acc: 0.397201, train avg acc: 0.193018, train iou: 0.889392
Test 1984, loss: 2.372624, test acc: 0.375116, test avg acc: 0.165928, test iou: 0.904366
Train 1985, loss: 2.348182, train acc: 0.387629, train avg acc: 0.190588, train iou: 0.885682
Test 1985, loss: 2.370390, test acc: 0.374345, test avg acc: 0.165133, test iou: 0.888351
Train 1986, loss: 2.317139, train acc: 0.397030, train avg acc: 0.193306, train iou: 0.889969
Test 1986, loss: 2.370961, test acc: 0.375983, test avg acc: 0.166822, test iou: 0.877985
Train 1987, loss: 2.316418, train acc: 0.397420, train avg acc: 0.195644, train iou: 0.942427
Test 1987, loss: 2.371313, test acc: 0.382780, test avg acc: 0.165983, test iou: 0.886651
Train 1988, loss: 2.317214, train acc: 0.396640, train avg acc: 0.192890, train iou: 0.942427
Test 1988, loss: 2.370719, test acc: 0.381412, test avg acc: 0.167635, test iou: 0.886651
Train 1989, loss: 2.316408, train acc: 0.396505, train avg acc: 0.195307, train iou: 0.875435
Test 1989, loss: 2.371716, test acc: 0.383513, test avg acc: 0.166739, test iou: 0.904366
Train 1990, loss: 2.315710, train acc: 0.396195, train avg acc: 0.190652, train iou: 0.888971
Test 1990, loss: 2.370887, test acc: 0.379709, test avg acc: 0.165399, test iou: 0.904366
Train 1991, loss: 2.324537, train acc: 0.396462, train avg acc: 0.195776, train iou: 0.842960
Test 1991, loss: 2.371117, test acc: 0.376484, test avg acc: 0.163988, test iou: 0.886651
Train 1992, loss: 2.316873, train acc: 0.397126, train avg acc: 0.195783, train iou: 0.885254
Test 1992, loss: 2.372163, test acc: 0.381046, test avg acc: 0.165371, test iou: 0.886651
Train 1993, loss: 2.316530, train acc: 0.396881, train avg acc: 0.193990, train iou: 0.911786
Test 1993, loss: 2.370981, test acc: 0.384566, test avg acc: 0.168906, test iou: 0.886651
Train 1994, loss: 2.317256, train acc: 0.397500, train avg acc: 0.195274, train iou: 0.885590
Test 1994, loss: 2.370783, test acc: 0.384566, test avg acc: 0.168906, test iou: 0.851485
Train 1995, loss: 2.317682, train acc: 0.396183, train avg acc: 0.194800, train iou: 0.842747
Test 1995, loss: 2.370464, test acc: 0.382645, test avg acc: 0.165876, test iou: 0.879140
Train 1996, loss: 2.321881, train acc: 0.395804, train avg acc: 0.196949, train iou: 0.890783
Test 1996, loss: 2.372171, test acc: 0.383969, test avg acc: 0.167209, test iou: 0.851485
Train 1997, loss: 2.318581, train acc: 0.396201, train avg acc: 0.191142, train iou: 0.942346
Test 1997, loss: 2.372860, test acc: 0.382697, test avg acc: 0.165897, test iou: 0.933891
Train 1998, loss: 2.317962, train acc: 0.397513, train avg acc: 0.192583, train iou: 0.885791
Test 1998, loss: 2.372269, test acc: 0.382337, test avg acc: 0.165531, test iou: 0.829975
Train 1999, loss: 2.317956, train acc: 0.395973, train avg acc: 0.192885, train iou: 0.911928
Test 1999, loss: 2.372949, test acc: 0.382523, test avg acc: 0.167770, test iou: 0.933891

[mmmmimic]

Over the next iterations, the loss remains around 2.37, the acc remains around 0.38, the avgacc remains around 0.17, and the iou fluctuates between 0.85 and 0.93. By the way, my dataset only has 164 samples. Test 1945, loss: 2.370986, test acc: 0.384566, test avg acc: 0.168906, test iou: 0.879140 Train 1946, loss: 2.317072, train acc: 0.397260, train avg acc: 0.195221, train iou: 0.889861 Test 1946, loss: 2.371344, test acc: 0.383127, test avg acc: 0.166341, test iou: 0.889699 Train 1947, loss: 2.315615, train acc: 0.396222, train avg acc: 0.192418, train iou: 0.889738 Test 1947, loss: 2.370504, test acc: 0.383436, test avg acc: 0.166659, test iou: 0.889699 Train 1948, loss: 2.315509, train acc: 0.397189, train avg acc: 0.193725, train iou: 0.889569 Test 1948, loss: 2.370401, test acc: 0.384566, test avg acc: 0.168906, test iou: 0.829975 Train 1949, loss: 2.316411, train acc: 0.396824, train avg acc: 0.192293, train iou: 0.885631 Test 1949, loss: 2.371268, test acc: 0.382980, test avg acc: 0.166189, test iou: 0.889699 Train 1950, loss: 2.320550, train acc: 0.396222, train avg acc: 0.193749, train iou: 0.912101 Test 1950, loss: 2.370480, test acc: 0.384566, test avg acc: 0.168906, test iou: 0.904366 Train 1951, loss: 2.320201, train acc: 0.396717, train avg acc: 0.196277, train iou: 0.842722 Test 1951, loss: 2.371929, test acc: 0.382620, test avg acc: 0.165818, test iou: 0.886651 Train 1952, loss: 2.318060, train acc: 0.396544, train avg acc: 0.193250, train iou: 0.942122 Test 1952, loss: 2.370918, test acc: 0.375765, test avg acc: 0.166597, test iou: 0.888351 Train 1953, loss: 2.318564, train acc: 0.396792, train avg acc: 0.193574, train iou: 0.889375 Test 1953, loss: 2.371295, test acc: 0.382755, test avg acc: 0.166494, test iou: 0.886651 Train 1954, loss: 2.317349, train acc: 0.397518, train avg acc: 0.193482, train iou: 0.889754 Test 1954, loss: 2.369916, test acc: 0.383217, test avg acc: 0.166434, test iou: 0.889699 Train 1955, loss: 2.318933, train acc: 0.395877, train avg acc: 0.191355, train iou: 0.873974 Test 1955, loss: 2.372064, test acc: 0.384566, test avg acc: 0.168906, test iou: 0.886651 Train 1956, loss: 2.316915, train acc: 0.396852, train avg acc: 0.192248, train iou: 0.842696 Test 1956, loss: 2.369206, test acc: 0.383391, test avg acc: 0.166613, test iou: 0.886651 Train 1957, loss: 2.320047, train acc: 0.396649, train avg acc: 0.193815, train iou: 0.889426 Test 1957, loss: 2.371206, test acc: 0.381964, test avg acc: 0.165144, test iou: 0.889699 Train 1958, loss: 2.318520, train acc: 0.396069, train avg acc: 0.194339, train iou: 0.942174 Test 1958, loss: 2.371605, test acc: 0.384566, test avg acc: 0.168906, test iou: 0.904366 Train 1959, loss: 2.317251, train acc: 0.397137, train avg acc: 0.193927, train iou: 0.911792 Test 1959, loss: 2.371200, test acc: 0.384566, test avg acc: 0.168906, test iou: 0.886651 Train 1960, loss: 2.321551, train acc: 0.395469, train avg acc: 0.195617, train iou: 0.906708 Test 1960, loss: 2.369627, test acc: 0.372616, test avg acc: 0.164244, test iou: 0.828898 Train 1961, loss: 2.319752, train acc: 0.397180, train avg acc: 0.192629, train iou: 0.889766 Test 1961, loss: 2.369780, test acc: 0.375212, test avg acc: 0.166027, test iou: 0.904366 Train 1962, loss: 2.316466, train acc: 0.397552, train avg acc: 0.192639, train iou: 0.911485 Test 1962, loss: 2.371272, test acc: 0.375662, test avg acc: 0.164019, test iou: 0.829310 Train 1963, loss: 2.317489, train acc: 0.396350, train avg acc: 0.191684, train iou: 0.889883 Test 1963, loss: 2.371849, test acc: 0.382639, test avg acc: 0.165838, test iou: 0.933891 Train 1964, loss: 2.321414, train acc: 0.396528, train avg acc: 0.195727, train iou: 0.911882 Test 1964, loss: 2.371795, test acc: 0.385273, test avg acc: 0.168554, test iou: 0.886651 Train 1965, loss: 2.323450, train acc: 0.397828, train avg acc: 0.193304, train iou: 0.911355 Test 1965, loss: 2.372183, test acc: 0.383551, test avg acc: 0.166778, test iou: 0.904366 Train 1966, loss: 2.318480, train acc: 0.395788, train avg acc: 0.191425, train iou: 0.942607 Test 1966, loss: 2.371025, test acc: 0.382729, test avg acc: 0.165930, test iou: 0.933891 Train 1967, loss: 2.317258, train acc: 0.395061, train avg acc: 0.190623, train iou: 0.889453 Test 1967, loss: 2.371001, test acc: 0.384464, test avg acc: 0.167719, test iou: 0.829975 Train 1968, loss: 2.318910, train acc: 0.395464, train avg acc: 0.192160, train iou: 0.889790 Test 1968, loss: 2.371105, test acc: 0.383583, test avg acc: 0.166811, test iou: 0.904366 Train 1969, loss: 2.319059, train acc: 0.396430, train avg acc: 0.193149, train iou: 0.875410 Test 1969, loss: 2.371892, test acc: 0.383384, test avg acc: 0.166606, test iou: 0.879140 Train 1970, loss: 2.321073, train acc: 0.397089, train avg acc: 0.195461, train iou: 0.911973 Test 1970, loss: 2.371132, test acc: 0.384566, test avg acc: 0.168906, test iou: 0.889699 Train 1971, loss: 2.316530, train acc: 0.397068, train avg acc: 0.193086, train iou: 0.890034 Test 1971, loss: 2.372177, test acc: 0.384566, test avg acc: 0.168906, test iou: 0.904366 Train 1972, loss: 2.319103, train acc: 0.396302, train avg acc: 0.191644, train iou: 0.817738 Test 1972, loss: 2.372295, test acc: 0.382973, test avg acc: 0.166182, test iou: 0.886651 Train 1973, loss: 2.317512, train acc: 0.397119, train avg acc: 0.191563, train iou: 0.912044 Test 1973, loss: 2.373002, test acc: 0.384412, test avg acc: 0.168706, test iou: 0.829975 Train 1974, loss: 2.319139, train acc: 0.394892, train avg acc: 0.192304, train iou: 0.889326 Test 1974, loss: 2.372499, test acc: 0.379144, test avg acc: 0.164934, test iou: 0.878635 Train 1975, loss: 2.319003, train acc: 0.396311, train avg acc: 0.193295, train iou: 0.885752 Test 1975, loss: 2.371416, test acc: 0.383513, test avg acc: 0.166739, test iou: 0.886651 Train 1976, loss: 2.318240, train acc: 0.396523, train avg acc: 0.194344, train iou: 0.911730 Test 1976, loss: 2.372317, test acc: 0.382851, test avg acc: 0.166056, test iou: 0.886651 Train 1977, loss: 2.322701, train acc: 0.395271, train avg acc: 0.193681, train iou: 0.839193 Test 1977, loss: 2.370505, test acc: 0.383795, test avg acc: 0.167030, test iou: 0.829975 Train 1978, loss: 2.316416, train acc: 0.396902, train avg acc: 0.192933, train iou: 0.885392 Test 1978, loss: 2.370879, test acc: 0.379035, test avg acc: 0.165498, test iou: 0.904366 Train 1979, loss: 2.316395, train acc: 0.396265, train avg acc: 0.192222, train iou: 0.889455 Test 1979, loss: 2.370659, test acc: 0.375032, test avg acc: 0.165842, test iou: 0.877985 Train 1980, loss: 2.316475, train acc: 0.396505, train avg acc: 0.193750, train iou: 0.885213 Test 1980, loss: 2.371453, test acc: 0.383699, test avg acc: 0.166931, test iou: 0.933891 Train 1981, loss: 2.315943, train acc: 0.396808, train avg acc: 0.191838, train iou: 0.889390 Test 1981, loss: 2.371143, test acc: 0.377332, test avg acc: 0.168214, test iou: 0.887617 Train 1982, loss: 2.316804, train acc: 0.396998, train avg acc: 0.193861, train iou: 0.942274 Test 1982, loss: 2.371442, test acc: 0.379594, test avg acc: 0.165903, test iou: 0.829517 Train 1983, loss: 2.317900, train acc: 0.396674, train avg acc: 0.193601, train iou: 0.889659 Test 1983, loss: 2.371251, test acc: 0.383282, test avg acc: 0.166500, test iou: 0.886651 Train 1984, loss: 2.315678, train acc: 0.397201, train avg acc: 0.193018, train iou: 0.889392 Test 1984, loss: 2.372624, test acc: 0.375116, test avg acc: 0.165928, test iou: 0.904366 Train 1985, loss: 2.348182, train acc: 0.387629, train avg acc: 0.190588, train iou: 0.885682 Test 1985, loss: 2.370390, test acc: 0.374345, test avg acc: 0.165133, test iou: 0.888351 Train 1986, loss: 2.317139, train acc: 0.397030, train avg acc: 0.193306, train iou: 0.889969 Test 1986, loss: 2.370961, test acc: 0.375983, test avg acc: 0.166822, test iou: 0.877985 Train 1987, loss: 2.316418, train acc: 0.397420, train avg acc: 0.195644, train iou: 0.942427 Test 1987, loss: 2.371313, test acc: 0.382780, test avg acc: 0.165983, test iou: 0.886651 Train 1988, loss: 2.317214, train acc: 0.396640, train avg acc: 0.192890, train iou: 0.942427 Test 1988, loss: 2.370719, test acc: 0.381412, test avg acc: 0.167635, test iou: 0.886651 Train 1989, loss: 2.316408, train acc: 0.396505, train avg acc: 0.195307, train iou: 0.875435 Test 1989, loss: 2.371716, test acc: 0.383513, test avg acc: 0.166739, test iou: 0.904366 Train 1990, loss: 2.315710, train acc: 0.396195, train avg acc: 0.190652, train iou: 0.888971 Test 1990, loss: 2.370887, test acc: 0.379709, test avg acc: 0.165399, test iou: 0.904366 Train 1991, loss: 2.324537, train acc: 0.396462, train avg acc: 0.195776, train iou: 0.842960 Test 1991, loss: 2.371117, test acc: 0.376484, test avg acc: 0.163988, test iou: 0.886651 Train 1992, loss: 2.316873, train acc: 0.397126, train avg acc: 0.195783, train iou: 0.885254 Test 1992, loss: 2.372163, test acc: 0.381046, test avg acc: 0.165371, test iou: 0.886651 Train 1993, loss: 2.316530, train acc: 0.396881, train avg acc: 0.193990, train iou: 0.911786 Test 1993, loss: 2.370981, test acc: 0.384566, test avg acc: 0.168906, test iou: 0.886651 Train 1994, loss: 2.317256, train acc: 0.397500, train avg acc: 0.195274, train iou: 0.885590 Test 1994, loss: 2.370783, test acc: 0.384566, test avg acc: 0.168906, test iou: 0.851485 Train 1995, loss: 2.317682, train acc: 0.396183, train avg acc: 0.194800, train iou: 0.842747 Test 1995, loss: 2.370464, test acc: 0.382645, test avg acc: 0.165876, test iou: 0.879140 Train 1996, loss: 2.321881, train acc: 0.395804, train avg acc: 0.196949, train iou: 0.890783 Test 1996, loss: 2.372171, test acc: 0.383969, test avg acc: 0.167209, test iou: 0.851485 Train 1997, loss: 2.318581, train acc: 0.396201, train avg acc: 0.191142, train iou: 0.942346 Test 1997, loss: 2.372860, test acc: 0.382697, test avg acc: 0.165897, test iou: 0.933891 Train 1998, loss: 2.317962, train acc: 0.397513, train avg acc: 0.192583, train iou: 0.885791 Test 1998, loss: 2.372269, test acc: 0.382337, test avg acc: 0.165531, test iou: 0.829975 Train 1999, loss: 2.317956, train acc: 0.395973, train avg acc: 0.192885, train iou: 0.911928 Test 1999, loss: 2.372949, test acc: 0.382523, test avg acc: 0.167770, test iou: 0.933891

Hi! This could be because the zero-division returning value is 1 in IoU computation. Specifically, in our implementation we followed dgcnn: if the union of ground truth and prediction points is empty, then count part IoU as 1. By modifying 'iou=1' to 'iou=0' in line 40 in main_partseg.py, the zero-division returning value should be 0 and by then IoU will not be 'overestimated'.

For the model performance, there could be multiple reasons. Could you try to tune the hyperparameters on your dataset, e.g., decreasing the model size and the dropout rate?