S4Former

KO | EN

gitlite — search

#react #ai #llm #go #golang #android #machine-learning #deep-learning #linux #macos #nodejs #kubernetes #openai #ios

(★ 17)

Training Vision Transformers for Semi-Supervised Semantic Segmentation

S4Former Latest Version Download

Download Latest Version (.zip)

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

## Introduction

This is the repo for our CVPR2024 paper "Training Vision Transformers for Semi-Supervised Semantic Segmentation".

## Setup

Please refer to [get_started.md](docs/en/get_started.md#installation) for MMSegmentation environments installation and [dataset_prepare.md](docs/en/dataset_prepare.md#prepare-datasets) for Pascal VOC 2012 and Cityscapes datasets preparation. Datasplits for labeled/unlabeled data division have been given under the `data` folder.

## Pre-Training Weight

#### 1. Download Pre-trained Weights

We recommend using the pre-trained weights provided by OpenMMLab, which are already adapted for MMCV/MMSegmentation.

- **DeiT-Base (Patch16, ImageNet-1k)**  
  Download link:  [deit-base_pt-16xb64_in1k_20220216-db63c16c.pth](https://download.openmmlab.com/mmclassification/v0/deit/deit-base_pt-16xb64_in1k_20220216-db63c16c.pth)

**After downloading, rename it to:**

```bash
deit_base_p16.pth
```

#### 2. Place the Weights

Put the weight file under the `pretrain/` directory:

~~~
project_root/
│── pretrain/
│    └── deit_base_p16.pth
│── configs/
│── mmseg/
│── tools/
~~~

#### 3. Convert the Weight Format (if needed)

In some cases, the key names of the pre-trained weights may not fully match the current implementation.
 If you encounter `unexpected key(s) in state_dict` or `missing key(s)` errors,
 please run the following script to convert the weight format:

~~~python
import torch

# 1. Load mmcls DeiT weights
ckpt = torch.load("pretrain/deit_base_p16.pth", map_location="cpu")
state_dict = ckpt.get("state_dict", ckpt)

new_state_dict = {}
for k, v in state_dict.items():
    new_k = k
    # remove "backbone." prefix
    if new_k.startswith("backbone."):
        new_k = new_k.replace("backbone.", "")
    # attention qkv mapping
    if "attn.qkv.weight" in new_k:
        new_k = new_k.replace("attn.qkv.weight", "attn.attn.in_proj_weight")
    if "attn.qkv.bias" in new_k:
        new_k = new_k.replace("attn.qkv.bias", "attn.attn.in_proj_bias")
    if "attn.proj.weight" in new_k:
        new_k = new_k.replace("attn.proj.weight", "attn.attn.out_proj.weight")
    if "attn.proj.bias" in new_k:
        new_k = new_k.replace("attn.proj.bias", "attn.attn.out_proj.bias")
    new_state_dict[new_k] = v

# 2. Save the converted weights
torch.save(new_state_dict, "pretrain/deit_base_p16.pth")
print("Converted and saved: pretrain/deit_base_p16.pth")
~~~

## Running Example
We run our S<sup>4</sup>Former based on the segmentation network of SegFormer([paper](https://arxiv.org/abs/2105.15203)|[project](https://github.com/open-mmlab/mmsegmentation/tree/master/configs/segformer)). The batch size is set to 8 for both labeled images and unlabeled images. Here we take 1/8 labeled data protocal on Pascal VOC 2012 as the example.

### Supervised-Only 
```bash
# use torch.distributed.launch
sh ./tools/dist_train.sh \
configs/setr/setr_deit-base_pup_bs_8_512x512_80k_pascal_1over16_split_classic_sup.py 2 \
--seed 1999 
```
### Mean Teacher Baseline
```bash
# use torch.distributed.launch
sh ./tools/dist_train.sh \
configs/setr/setr_deit-base_pup_bs_8_512x512_80k_pascal_1over16_split_classic_semi_beta_1_th_0.95_MT.py 2 \
--seed 1999 
```
### S<sup>4</sup>Former (Ours)
```bash
# use torch.distributed.launch
sh ./tools/dist_train.sh \
configs/setr/setr_deit-base_pup_bs_8_512x512_80k_pascal_1over16_split_classic_semi_beta_1_th_0.95_MT_w_ours.py 2 \
--seed 1999 
```

We thank so much for the feedbacks and updates made by Shun Zuo.

S4Former

S4Former Latest Version Download

Today's picks

Was this content helpful?

Recent Feedback

S4Former

S4Former Latest Version Download

Today's picks

Was this content helpful?

Recent Feedback

Download README