Skip to content
View baptistepras's full-sized avatar
🎯
Focusing
🎯
Focusing
2 followers · 2 following

Achievements

Achievement: Pull Shark

Achievements

Achievement: Pull Shark

Highlights

  • Pro

Block or report baptistepras

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
baptistepras/README.md

Hi, I'm Baptiste Pras

Master's student in Artificial Intelligence at Université Paris-Saclay (PhD track), with a funded PhD position at Université Paris-Saclay lined up after the master, on robust and explainable multimodal detection of small aerial targets from radar and infrared signals.

My research covers three areas:

  • Computer vision: object counting, detection, video prediction;
  • NLP: biomedical entity linking, information retrieval, summarization;
  • Imbalanced learning: how class ratios in the training data shape what classifiers learn.

Publications

  • Point-Based Counting of Cereals and Legumes in Intercropped Fields
    B. Pras. Junior Conference on Data Science and Engineering (JDSE), 2026, poster. HAL · code
  • Fine-Grained Mention-Level Analysis of Biomedical Entity Linking Models
    B. Pras and N. Naderi. Medical Informatics Europe (MIE), 2026. Paper · code
  • Revisiting Optimal Class Ratios in Imbalanced Learning
    B. Pras. Junior Conference on Data Science and Engineering (JDSE), 2025. HAL · code

Tools

Python PyTorch scikit-learn Hugging Face NumPy OpenCV C++ Java OCaml Linux Slurm LaTeX

Contact

Website · LinkedIn · Google Scholar · ORCID

Pinned Loading

  1. Structure-detection-in-fusion-plasma-simulations Structure-detection-in-fusion-plasma-simulations Public

    YOLOv8 with pseudo-labeling and a learned post-filter to detect blobs in Tokam2D fusion plasma simulations. 4th/94 on Codabench, 81% AP50.

    Python

  2. peekaboo-prednet peekaboo-prednet Public

    Keeping track of hidden objects with predictive coding networks. A controlled occlusion benchmark for PredNet: tracking through occlusion, recovery after reappearance, and surprise.

    Python

  3. Point-Based-Counting-of-Cereals-and-Legumes-in-Intercropped-Fields Point-Based-Counting-of-Cereals-and-Legumes-in-Intercropped-Fields Public

    Per-species counting of wheat and pea in intercrops with the PET point-query transformer. 4.7% MAPE on wheat tips, 7.8% on pea. JDSE 2026 poster.

    Python

  4. CycleGAN CycleGAN Public

    CycleGAN (Zhu et al., ICCV 2017) rewritten from scratch in pure NumPy with hand-written backward passes. Test cycle L1 of 0.20 on horse2zebra and apple2orange.

    Python

  5. financial-news-impact-prediction financial-news-impact-prediction Public

    Frugal NLP pipeline turning financial news into (date, ticker, impact) events: map-reduce summarization with Flan-T5, LLM-as-a-judge audit, NER ticker linking.

    Python

  6. Scientific-Article-Information-Retrieval Scientific-Article-Information-Retrieval Public

    Citation retrieval on 20k papers: BM25, dense encoders, citation context mining, and XGBoost learning to rank. MAP from 0.45 to 0.67.

    Python