billpsomas/simpool

This repo contains the official implementation of ICCV 2023 paper "Keep It SimPool: Who Said Supervised Transformers Suffer from Attention Deficit?"

29
/ 100
Experimental

This project helps computer vision researchers and practitioners generate high-quality attention maps from their image classification models. By replacing the default pooling mechanism in convolutional networks or vision transformers, it takes an image as input and outputs a classified image alongside detailed attention maps that clearly delineate object boundaries. Anyone working with image recognition, especially those fine-tuning or interpreting deep learning models, would find this useful.

101 stars. No commits in the last 6 months.

Use this if you need to visualize exactly which parts of an image your deep learning model is focusing on to make a classification, and you require highly accurate attention maps that outline objects precisely.

Not ideal if you are looking for a pre-trained, off-the-shelf image classification solution without needing to delve into model architecture modifications or interpretability features.

image-classification deep-learning-interpretability computer-vision-research model-analysis
Stale 6m No Package No Dependents
Maintenance 0 / 25
Adoption 9 / 25
Maturity 16 / 25
Community 4 / 25

How are scores calculated?

Stars

101

Forks

2

Language

Python

License

Apache-2.0

Last pushed

Dec 05, 2023

Commits (30d)

0

Get this data via API

curl "https://pt-edge.onrender.com/api/v1/quality/ml-frameworks/billpsomas/simpool"

Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.