OWSOL

Code of our paper Open-World Weakly-Supervised Object Localization.

Dependencies

• Python 3
• PyTorch 1.7.1
• OpenCV-Python
• Numpy
• Scipy
• MatplotLib
• faiss-gpu
• munch

Dataset

ImageNet-1K

"train" and "val" splits of original ImageNet are treated as our train and test, ImageNetV2 is treated as our val.

Make sure your dataset/ILSVRC folder is structured as follows:

├── ILSVRC/
|   ├── train/
|	|	|── n01440764
|	|	|── n01443537
|	|	|── ...
|   ├── val
|	|	|── n01440764
|	|	|── n01443537
|	|	|── ...
|   ├── val2
|	|	|── 0
|	|	|── 1
|   |	└── ...

Metadata

You can download the annotations of datasets from Download metadata, the password is dqdf.

Make sure your metadata/ImageNet folder is structured as follows:

├── ImageNet/
|   ├── train/
|	|	|── class_labels.txt
|	|	|── image_ids.txt
|	|	|── image_ids_labeled.txt
|   ├── test
|	|	|── class_labels.txt
|	|	|── image_ids.txt
|	|	|── image_sizes.txt
|	|	|── localization.txt.txt
|	|	|── partitions.txt
|   ├── val
|	|	|── class_labels.txt
|	|	|── image_ids.txt
|	|	|── image_sizes.txt
|	|	|── localization.txt.txt
|	|	|── partitions.txt

for test and val, the partitions of Known, Nov-S and Nov-D is described in partitions.txt. In detail, 0, 1 and 2 are correspond to Known, Nov-S and Nov-D, respectively.

Training

perform contrastive representation co-learning on ImageNet-1K dataset

bash bash_script/train_imagenet.sh

G-CAM

perform g-cam on ImageNet-1K dataset

bash bash_script/gcam_imagenet.sh