paulgavrikov/CNN-Filter-DB
A database of over 1.4 billion 3x3 convolution filters extracted from hundreds of diverse CNN models with relevant meta information (CVPR 2022 ORAL)
This project provides a comprehensive database of 1.4 billion 3x3 convolution filters collected from numerous trained Convolutional Neural Networks. It allows researchers to investigate how design choices like dataset, architecture, or task affect the learned filter weights, helping to understand why some models transfer or fine-tune better than others. Computer vision researchers and deep learning practitioners can use this to gain insights into model robustness and transferability.
No commits in the last 6 months.
Use this if you are a computer vision researcher or deep learning practitioner seeking to understand the underlying properties of trained CNN filters and how they influence model performance, robustness, or transferability across different applications.
Not ideal if you are looking for a pre-trained model to use directly for a specific computer vision task, as this project provides a dataset for analysis rather than an out-of-the-box solution.
Stars
34
Forks
3
Language
Jupyter Notebook
License
CC-BY-SA-4.0
Category
Last pushed
Jun 28, 2023
Commits (30d)
0
Get this data via API
curl "https://pt-edge.onrender.com/api/v1/quality/ml-frameworks/paulgavrikov/CNN-Filter-DB"
Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.
Higher-rated alternatives
Jittor/jittor
Jittor is a high-performance deep learning framework based on JIT compiling and meta-operators.
zhanghang1989/ResNeSt
ResNeSt: Split-Attention Networks
berniwal/swin-transformer-pytorch
Implementation of the Swin Transformer in PyTorch.
NVlabs/FasterViT
[ICLR 2024] Official PyTorch implementation of FasterViT: Fast Vision Transformers with...
ViTAE-Transformer/ViTPose
The official repo for [NeurIPS'22] "ViTPose: Simple Vision Transformer Baselines for Human Pose...