Search Results for "protocol test harness"

Torchreid is a library for deep-learning person re-identification, written in PyTorch and developed for our ICCV’19 project, Omni-Scale Feature Learning for Person Re-Identification. In "deep-person-reid/scripts/", we provide a unified interface to train and test a model. See "scripts/main.py" and "scripts/default_config.py" for more details. The folder "configs/" contains some predefined configs which you can use as a starting point. The code will automatically (download and) load the...

...The repository provides training and evaluation code for standard few-shot benchmarks such as miniImageNet and Omniglot, making it possible to reproduce the experimental results reported in the paper. It includes model definitions, data loading logic, episodic training loops, and scripts that implement the N-way K-shot evaluation protocol common in few-shot research. Researchers can use this codebase as a starting point to test new ideas, modify relation modules, or transfer the approach to new datasets.