# Debiased Mapping for Full-Reference Image Quality Assessment
This is the repository of paper [Debiased Mapping for Full-Reference Image Quality Assessment](https://arxiv.org/abs/2302.11464).
### Highlights:
* The perception bias of existing deep-feature based FR-IQA measures is explored, which may cause inferior performance on misaligned features and restored content.
* We propose an SVD-based debiased mapping to mitigate the perception bias. Specifically, the SVs distance and base angle consistency are designed to capture and measure the feature distortion reliably.
### ====== PyTorch Implementation ======
**Installation:**
- ```pip install DMM-PyTorch```
**Requirements:**
- Python >= 3.6
- PyTorch >= 1.0
**Usage:**
```python
from DMM_PyTorch import DMM
from torchvision import transforms
from PIL import Image
def prepare_PIL_Image(PIL_Image):
msize = min(PIL_Image.size)
if msize>128:
tar_size = max(int(msize/(1.0*48))*32,128)
image =transforms.functional.resize(PIL_Image,tar_size)
image = transforms.ToTensor()(image)
return image.unsqueeze(0)
model = DMM().cuda()
ref_pth = './Images/I04.BMP'
dist_pth = './Images/i04_24_2.bmp'
ref = prepare_PIL_Image(Image.open(ref_pth).convert("RGB")).cuda()
dist = prepare_PIL_Image(Image.open(dist_pth).convert("RGB")).cuda()
dmm_score = model(ref, dist)
print(dmm_score)
```
or
```bash
git clone https://github.com/Baoliang93/DMM
cd DMM_PyTorch
python DMM.py --ref <ref_path> --dist <dist_path>
```
Raw data
{
"_id": null,
"home_page": "https://github.com/Baoliang93/DMM",
"name": "DMM-PyTorch",
"maintainer": null,
"docs_url": null,
"requires_python": null,
"maintainer_email": null,
"keywords": "pytorch, Quality, Debiased, Full-reference, metric",
"author": "Baoliang CHEN",
"author_email": "blchen@scnu.edu.cn",
"download_url": "https://files.pythonhosted.org/packages/b2/89/0c22052f021c89602adc78d75aebee4dbed171f5f6d93d968f29e7521d6c/DMM_PyTorch-0.1.tar.gz",
"platform": "python",
"description": "# Debiased Mapping for Full-Reference Image Quality Assessment\r\n\r\nThis is the repository of paper [Debiased Mapping for Full-Reference Image Quality Assessment](https://arxiv.org/abs/2302.11464). \r\n\r\n### Highlights:\r\n* The perception bias of existing deep-feature based FR-IQA measures is explored, which may cause inferior performance on misaligned features and restored content.\r\n* We propose an SVD-based debiased mapping to mitigate the perception bias. Specifically, the SVs distance and base angle consistency are designed to capture and measure the feature distortion reliably.\r\n\r\n\r\n### ====== PyTorch Implementation ======\r\n**Installation:** \r\n- ```pip install DMM-PyTorch```\r\n\r\n**Requirements:** \r\n- Python >= 3.6\r\n- PyTorch >= 1.0\r\n\r\n**Usage:** \r\n```python\r\n\r\nfrom DMM_PyTorch import DMM\r\nfrom torchvision import transforms\r\nfrom PIL import Image\r\n\r\ndef prepare_PIL_Image(PIL_Image):\r\n msize = min(PIL_Image.size)\r\n if msize>128:\r\n tar_size = max(int(msize/(1.0*48))*32,128)\r\n image =transforms.functional.resize(PIL_Image,tar_size)\r\n image = transforms.ToTensor()(image)\r\n return image.unsqueeze(0)\r\n\r\nmodel = DMM().cuda()\r\n\r\nref_pth = './Images/I04.BMP' \r\ndist_pth = './Images/i04_24_2.bmp' \r\n \r\nref = prepare_PIL_Image(Image.open(ref_pth).convert(\"RGB\")).cuda()\r\ndist = prepare_PIL_Image(Image.open(dist_pth).convert(\"RGB\")).cuda()\r\n\r\ndmm_score = model(ref, dist)\r\nprint(dmm_score)\r\n```\r\nor\r\n\r\n```bash\r\ngit clone https://github.com/Baoliang93/DMM\r\ncd DMM_PyTorch\r\npython DMM.py --ref <ref_path> --dist <dist_path>\r\n```\r\n\r\n\r\n\r\n",
"bugtrack_url": null,
"license": "MIT",
"summary": "Debiased Mapping for Full-Reference Image Quality Assessment",
"version": "0.1",
"project_urls": {
"Homepage": "https://github.com/Baoliang93/DMM"
},
"split_keywords": [
"pytorch",
" quality",
" debiased",
" full-reference",
" metric"
],
"urls": [
{
"comment_text": "",
"digests": {
"blake2b_256": "cb306f7749fee152bae2303803a6738e5a69afaec65ca633964dfff573aa3cfe",
"md5": "b924bfb7ede2e3431fada8ecd390a4d7",
"sha256": "f66ff90a95d2b548738c6288720300db327f2c0a18f7c90bb6c0c5c7d7a493e3"
},
"downloads": -1,
"filename": "DMM_PyTorch-0.1-py3-none-any.whl",
"has_sig": false,
"md5_digest": "b924bfb7ede2e3431fada8ecd390a4d7",
"packagetype": "bdist_wheel",
"python_version": "py3",
"requires_python": null,
"size": 5480,
"upload_time": "2024-09-16T13:01:01",
"upload_time_iso_8601": "2024-09-16T13:01:01.584743Z",
"url": "https://files.pythonhosted.org/packages/cb/30/6f7749fee152bae2303803a6738e5a69afaec65ca633964dfff573aa3cfe/DMM_PyTorch-0.1-py3-none-any.whl",
"yanked": false,
"yanked_reason": null
},
{
"comment_text": "",
"digests": {
"blake2b_256": "b2890c22052f021c89602adc78d75aebee4dbed171f5f6d93d968f29e7521d6c",
"md5": "d337ff076e1a47718eb022789a84db7c",
"sha256": "80b1ba071a0cead87eb65e55906c11206ecdc217d6a5653dfc2df117f20fef5f"
},
"downloads": -1,
"filename": "DMM_PyTorch-0.1.tar.gz",
"has_sig": false,
"md5_digest": "d337ff076e1a47718eb022789a84db7c",
"packagetype": "sdist",
"python_version": "source",
"requires_python": null,
"size": 5174,
"upload_time": "2024-09-16T13:01:03",
"upload_time_iso_8601": "2024-09-16T13:01:03.160399Z",
"url": "https://files.pythonhosted.org/packages/b2/89/0c22052f021c89602adc78d75aebee4dbed171f5f6d93d968f29e7521d6c/DMM_PyTorch-0.1.tar.gz",
"yanked": false,
"yanked_reason": null
}
],
"upload_time": "2024-09-16 13:01:03",
"github": true,
"gitlab": false,
"bitbucket": false,
"codeberg": false,
"github_user": "Baoliang93",
"github_project": "DMM",
"travis_ci": false,
"coveralls": false,
"github_actions": false,
"lcname": "dmm-pytorch"
}
JSON
