CAMMA-public / rendezvous

A transformer-inspired neural network for surgical action triplet recognition from laparoscopic videos.
Other
24 stars 8 forks source link

Implementation of random scaling #15

Closed csEylLee closed 1 year ago

csEylLee commented 1 year ago

In your paper you writed:"We employed random scaling [0.5, 1.5]", but there is no such method in the released code. Can you provide the implementation of random scaling? I tried to code a random scale function and followed all the hyperparameter settings in the paper, but I still could not get the results in the paper.

csEylLee commented 1 year ago

Mean AP: I | V | T | IV | IT | IVT :::::: : 0.8274 | 0.5316 | 0.3626 | 0.3070 | 0.2689 | 0.2105

nwoyecid commented 1 year ago

This repo is a re-implementation of the original RDV. It does not exactly use everything as in the RDV paper, e.g. random scaling. If you implement it, you can make a pull request to contribute your implementation to this repo.