search

Tahlor / simple_hwr

7 stars 1 forks source link

TRACE ICDAR paper inference results #1

Open mok33 opened 1 year ago

mok33 commented 1 year ago

Hello, To my understanding, this repo contains the official implementation of "TRACE: A Differentiable Approach to Line-level Stroke Recovery for Offline Handwritten Text". Can you provide a zip of the recovered inks for IAM? If you can also provide the trained model that reproduces the results. It can be of great use for me, as I want to compare your approach with other SoTA methods.

Tahlor commented 1 year ago

Unfortunately, I don't have the results anymore, but you should be able to use the model from the example_weights folder. The stroke_recovery_offline.py should produce the results for you, otherwise, I might be able to look at it in a few weeks. Best of luck!

Alma-hanif commented 1 year ago

Hello, It's good to see your response here, Can you please also tell me which Config file did you use for the "TRACE: A Differentiable Approach to Line-level Stroke Recovery for Offline Handwritten Text" as there are so many different versions of config files in the config folder and I am not sure which one is for the TRACE approach.

Tahlor commented 1 year ago

The config in the example_weights folder should work with those weights. Excited to see your new work!