Person Search Term Project
- numpy>=1.18.1
- opencv>=3.4.2
- pytorch>=1.4.0
- torchvision>=0.5
- scikit-learn>=0.22.1
- Download CUHK-SYSU dataset and unzip it in person_search/dataset.
- Run
simclr_training.py
for representation learning.
- Run
PersonSearch.py
for person search.
- Dataset folder tree
person_search
|
+-- dataset
|
+-- CUHK-SYSU
|
+-- Image
| |
| +-- SSM
| +-- bbox (to be appeared by data_preprocessing.py)
| |
| +-- TestG50
| +-- Train_only1
|
+-- annotation
| |
| +-- Images.mat
| +-- Person.mat
| +-- pool.mat
| +-- test
| |
| +-- train_test
| |
| +-- TestG50.mat
|
+-- processed (to be appeared by data_preprocessing.py)
|
+-- images.csv
+-- TestG50_data.csv
- Code tree
person_search
|
+-- custom_data
| |
| +-- dataset.py (our own implementation)
| +-- data_preprocessing.py (our own implementation)
| +-- contrastive_learning.py
| +-- view_generator.py
| +-- gaussian_blur.py
|
+-- models
| |
| +-- pretrained_model.py
| +-- resnet_simclr.py
| +-- simclr.py
|
+-- simclr_utils
+-- torchvision_utils
+-- vision
+-- checkpoints (to be appeared during training step)
+-- results (to be appeared during testing step)
|
+-- simclr_training.py
+-- PersonSearch.py