DEVAI / instances /13_Style_Transfer_Perceptual_Loss_CustomImages_DL.json
DEVAI-benchmark's picture
Upload 55 files
6822471 verified
{
"name": "13_Style_Transfer_Perceptual_Loss_CustomImages_DL",
"query": "Please create a PyTorch Perceptual Loss project for image style transfer (refer to this paper: https://arxiv.org/pdf/1603.08155). You can build the Perceptual Loss Network using VGG16 in `src/model.py`. The project should combine content and style images, allow smooth adjustment of style intensity by tuning the weights of style loss and content loss, and save the stylized images in `results/figures/`. Additionally, log the processing time to `results/processing_time.txt`, and save the intermediate results of the style transfer process to `results/figures/intermediate_results.png`. For testing, input a famous content image (Mona Lisa) from [this link](https://upload.wikimedia.org/wikipedia/commons/thumb/e/ec/Mona_Lisa%2C_by_Leonardo_da_Vinci%2C_from_C2RMF_retouched.jpg/768px-Mona_Lisa%2C_by_Leonardo_da_Vinci%2C_from_C2RMF_retouched.jpg) and a famous style image (The Starry Night) from [this link](https://ia904705.us.archive.org/34/items/the-starry-night-vincent-van-gogh/The%20Starry%20Night%20-%20Vincent%20van%20Gogh/%21PREVIEW%21%20-%20The%20Starry%20Night%20-%20Vincent%20van%20Gogh.jpg), and generate a style-transfered image. Save the content, style, and style-transfered images to `data/content.jpg`, `data/style.jpg`, and `results/figures/`, respectively. The project should efficiently handle high-resolution images without excessive processing time.",
"tags": [
"Computer Vision",
"Generative Models"
],
"requirements": [
{
"requirement_id": 0,
"prerequisites": [],
"criteria": "A famous content image is inputted for testing, downloaded from [this link](https://upload.wikimedia.org/wikipedia/commons/thumb/e/ec/Mona_Lisa%2C_by_Leonardo_da_Vinci%2C_from_C2RMF_retouched.jpg/768px-Mona_Lisa%2C_by_Leonardo_da_Vinci%2C_from_C2RMF_retouched.jpg) and saved to `data/content.jpg`.",
"category": "Dataset or Environment",
"satisfied": null
},
{
"requirement_id": 1,
"prerequisites": [],
"criteria": "A famous style image is inputted for testing, downloaded from [this link](https://ia904705.us.archive.org/34/items/the-starry-night-vincent-van-gogh/The%20Starry%20Night%20-%20Vincent%20van%20Gogh/%21PREVIEW%21%20-%20The%20Starry%20Night%20-%20Vincent%20van%20Gogh.jpg) and saved in `data/style.jpg`.",
"category": "Dataset or Environment",
"satisfied": null
},
{
"requirement_id": 2,
"prerequisites": [],
"criteria": "The Perceptual loss model implemented in \"PyTorch\" is loaded in `src/model.py`.",
"category": "Machine Learning Method",
"satisfied": null
},
{
"requirement_id": 3,
"prerequisites": [
0,
1,
2
],
"criteria": "Stylized images are saved to the specified folder `results/figures/`.",
"category": "Other",
"satisfied": null
},
{
"requirement_id": 4,
"prerequisites": [
0,
1,
2
],
"criteria": "Style intensity is adjusted by tuning the weights of style loss and content loss in `src/model.py`.",
"category": "Machine Learning Method",
"satisfied": null
},
{
"requirement_id": 5,
"prerequisites": [
0,
1,
2,
3,
4
],
"criteria": "Processing time is recorded and saved as `results/processing_time.txt`.",
"category": "Performance Metrics",
"satisfied": null
},
{
"requirement_id": 6,
"prerequisites": [
0,
1,
2,
3,
4
],
"criteria": "Intermediate results of style transfer are saved as `results/figures/intermediate_results.png`.",
"category": "Visualization",
"satisfied": null
}
],
"preferences": [
{
"preference_id": 0,
"criteria": "The style transfer process should allow for smooth adjustment of style intensity, making the stylized image visually appealing.",
"satisfied": null
},
{
"preference_id": 1,
"criteria": "The project should handle high-resolution images efficiently without excessive processing time.",
"satisfied": null
}
],
"is_kaggle_api_needed": false,
"is_training_needed": false,
"is_web_navigation_needed": false,
"hint": "VGG16 was not originally designed for style transfer. However, the user's query states, 'Please create a PyTorch project for image style transfer using a pre-trained VGG16 model.' Ideally, a well-informed agent should create or find a model for style transfer networks that incorporate pre-trained VGG16, rather than simply loading the VGG16 model."
}