KO | EN

GitLite

Minimalist Open Source Repository Search Engine

gitlite — search

#dsh #dsh-plugin #ai-agent #developer-tools #codex #typescript #cli #deepseek-harness #claude-code #ai-agents #mcp #rust

LPCAM

★ 31 Open GitHub ↗

The official code of CVPR 2023 paper (Extracting Class Activation Maps from Non-Discriminative Features as well).

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

# LPCAM
The official code of CVPR 2023 paper (Extracting Class Activation Maps from Non-Discriminative Features as well).
[arXiv](https://arxiv.org/abs/2303.10334)

## Prerequisite
- Python 3.6, PyTorch 1.9, and others in environment.yml
- You can create the environment from environment.yml file
```
conda env create -f environment.yml
```
## Usage (PASCAL VOC)
### Step 1. Prepare dataset.
- Download PASCAL VOC 2012 devkit from [official website](http://host.robots.ox.ac.uk/pascal/VOC/voc2012/#devkit). [Download](http://host.robots.ox.ac.uk/pascal/VOC/voc2012/VOCtrainval_11-May-2012.tar). 
- You need to specify the path ('voc12_root') of your downloaded devkit in the following steps.
### Step 2. Train classification network and generate LPCAM seeds.
- Please specify a workspace to save the model and logs.
```
CUDA_VISIBLE_DEVICES=0 python run_sample.py --voc12_root ./VOCdevkit/VOC2012/ --work_space YOUR_WORK_SPACE --train_cam_pass True --make_cam_pass True --make_lpcam_pass True --eval_cam_pass True 
```
### Step 3. Train IRN and generate pseudo masks.
```
CUDA_VISIBLE_DEVICES=0 python run_sample.py --voc12_root ./VOCdevkit/VOC2012/ --work_space YOUR_WORK_SPACE --cam_to_ir_label_pass True --train_irn_pass True --make_sem_seg_pass True --eval_sem_seg_pass True 
```
You can download the pseudo labels from this [link](https://drive.google.com/file/d/1eDyD1b0nVI2WY4TcuN71yx8OoAOfsIlU/view?usp=share_link).

### Step 4. Train semantic segmentation network.
To train DeepLab-v2, we refer to [deeplab-pytorch](https://github.com/kazuto1011/deeplab-pytorch). 
We use the [ImageNet pre-trained model](https://drive.google.com/file/d/14soMKDnIZ_crXQTlol9sNHVPozcQQpMn/view?usp=sharing) for DeepLabV2 provided by [AdvCAM](https://github.com/jbeomlee93/AdvCAM).
Please replace the groundtruth masks with generated pseudo masks.

## Usage (MS COCO)
### Step 1. Prepare dataset.
- Download MS COCO images from the [official COCO website](https://cocodataset.org/#download).
- Generate mask from annotations (annToMask.py file in `./mscoco/`).
- Download MS COCO image-level labels provided by [ReCAM](https://github.com/zhaozhengChen/ReCAM) from [here](https://drive.google.com/drive/folders/1XCu51bAUK3nOvO-VVKD7kE9bIFpAECBR?usp=sharing) and put them in `./mscoco/`
### Step 2. Train classification network and generate LPCAM seeds.
- Please specify a workspace to save the model and logs.
```
CUDA_VISIBLE_DEVICES=0 python run_sample_coco.py --mscoco_root ../MSCOCO/ --work_space YOUR_WORK_SPACE --train_cam_pass True --make_cam_pass True --make_lpcam_pass True --eval_cam_pass True 
```
### Step 3. Train IRN and generate pseudo masks.
```
CUDA_VISIBLE_DEVICES=0 python run_sample_coco.py --mscoco_root ../MSCOCO/ --work_space YOUR_WORK_SPACE --cam_to_ir_label_pass True --train_irn_pass True --make_sem_seg_pass True --eval_sem_seg_pass True 
```
You can download the pseudo labels from this [link](https://drive.google.com/file/d/1UHglmht8PjJdRz71IrG2b8ENMN2Nn6P0/view?usp=share_link).

### Step 4. Train semantic segmentation network.
- The same as PASCAL VOC.

## Acknowledgment
This code is borrowed from [IRN](https://github.com/jiwoon-ahn/irn) and [ReCAM](https://github.com/zhaozhengChen/ReCAM).

LPCAM

Explore Similar Repositories

Was this content helpful?

Recent Feedback

Download README