hard-negative-mining 开源项目推荐（35 个）

ZhaoJ9014 /face.evoLVe

🔥🔥High-Performance Face Recognition Library on PaddlePaddle & PyTorch🔥🔥

junxia97 /ProGCL

[ICML 2022] "ProGCL: Rethinking Hard Negative Mining in Graph Contrastive Learning"

songweige /Contrastive-Learning-with-Non-Semantic-Negatives

Robust Contrastive Learning Using Negative Samples with Diminished Semantics (NeurIPS 2021)

DaoyuanLi2816 /labelbank

Retrieve + rerank over a closed label bank: LLM bi-encoders with self-mined hard negatives and a generative listwise reranker. Generalized from a silver-medal solution to Kaggle Eedi — Mining Misconceptions in Mathematics.

OpenMatch /ANCE-Tele

Code and data of the EMNLP 2022 Main Conference paper "Reduce Catastrophic Forgetting of Dense Retrieval Training with Teleportation Negatives".

saimj7 /Object-Detection-Algorithms

Detection algorithms and applications from famous papers; simple theory; solid code.

mala-lab /AUGCL

Official code for TNNLS paper "Affinity Uncertainty-based Hard Negative Mining in Graph Contrastive Learning"

bupt-ai-cz /ImageRetrieval_DF_CDVS

https://teacher.bupt.edu.cn/zhuchuang/en/index.htm

usbt0p /product-matching-hackudc-inditex

Winner project of the HackUDC 2026 Inditex AI challenge. From a photo of an outfit and given the Zara catalog, identify all the individual pieces the model is wearing.

Adam-Researchh /mlx-embed-finetune

Fine-tune single-vector embeddings on Apple Silicon with MLX: LoRA/QLoRA, GradCache, multi-positive training and teacher ranking distillation.

ntphuc149 /ViIR

ViIR: The Unified Framework for Fine-tuning Vietnamese Information Retrieval Models with Various Tuning Statergies.

shah-deven /SVM-Object-Detection

Train a SVM and for detecting human upper bodies in TV series The Big Bang Theory

thdwhtls /multimodal-grounding-posttraining

A visual grounding system for RefCOCO that freezes GroundingDINO candidates and trains a lightweight multimodal fusion reranker for region-text alignment and verifiable candidate ranking.

viniavskyi-ostap /superglue-lightning

SuperGlue training pipeline written using Pytorch Lightning

LohitMoharana /DeepOcean-AI

Computer vision pipeline for underwater object detection (mines, UUVs, divers) — custom YOLOv8 training with physics-based synthetic data generation and iterative hard-negative mining.

meng1994412 /Custom_Object_Detector

Custom Object Detector

IamOumarIbrahim /ieee-aiot

Hard-negative curation and sampling benchmark for mitigating false alarms in continuous edge AIoT perception.

stkef /matchlens

Finds the same product across marketplace listings from titles and photos. Built as a measured experiment ladder: BM25 → photo hashes → embeddings → learned fusion → fine-tuned models. Test F1 0.80 (BM25 0.68).

Supakorn289 /fire-smoke-detection-r3e6

End-to-end YOLOv8 Fire & Smoke Detection research: dataset curation, hard-negative mining, R1→R3 training, locked final test, ONNX deployment, and downloadable model release.

KashifMoin1410 /HNM-Based-Biomedical-Relation-Extraction

Implements relation extraction for biomedical texts using Hard Negative Mining to improve accuracy in identifying complex entity relationships. Includes code for data processing, training, and evaluation with BioC-format datasets.