winycg / anchor_net

Localizing Semantic Patches for Accelerating Image Classification (Updated 2 years, 11 months ago)

  • Public
  • 222 runs
  • GitHub
  • Paper

Input

input_image
*file

Image to be classified

string

Final-probs-plot: Plots the final prediction of top 10 classes Json: Get the final prediction of top 10 classes in json format Predictions-plot: Plot intermediate predictions of patches CAM-plot: Plot the intermediate CAM map from AnchorNet

Default: "Predictions-plot"

string
Shift + Return to add a new line

Comma separated list of floats where the ith float determines the allowed IOU to a next (i+1) patch proposal (6 max). Leave empty for a sinle patch proposal

Default: ""

Output

plot
Generated in

This output was created using a different version of the model, winycg/anchor_net:8f850419.

Run time and cost

This model costs approximately $0.053 to run on Replicate, or 18 runs per $1, but this varies depending on your inputs. It is also open source and you can run it on your own computer with Docker.

This model runs on Nvidia T4 GPU hardware. Predictions typically complete within 4 minutes. The predict time for this model varies significantly based on the inputs.

Readme

This project provides source code for the paper “Localizing Semantic Patches for Accelerating Image Classification.”

It introduce a two-step image classification pipeline: Given an image a lightweight CNN proposes k relevant patches then, the each patch is classified by a stronger CNN and the probabilities are gradually aggregated to a final prediciton

@inproceedings{yang2022localizing, title={Localizing Semantic Patches for Accelerating Image Classification}, author={Chuanguang Yang, Zhulin An, Yongjun Xu}, booktitle={IEEE International Conference on Multimedia and Expo}, year={2022} }