chenjshnn / Object-Detection-for-Graphical-User-Interface

Object Detection for Graphical User Interface: Old Fashioned or Deep Learning or a Combination?
Apache License 2.0
119 stars 27 forks source link

Request for dataset #3

Closed thuyhoang-hvtt closed 3 years ago

thuyhoang-hvtt commented 3 years ago

Hi Chen, I'm having trouble reproducing your results by running train.py with the rico dataset that I customized by myself but the results turn out not so good cuz the training loss always about 10. Could you provide me the dataset that you customized. Thanks in advance, Regards

chenjshnn commented 3 years ago

Hi,

This is my rico train/val/test dataset (without text elements). You can try to train the model on it to see if you can obtain a similar results. https://www.dropbox.com/sh/4higxc7e15abf1u/AACilquN6CZzadSymf44WAjGa?dl=0

Best, Jieshan

chenjshnn commented 3 years ago

If you still have a similar problem, please feel free to post an issue and email me (jieshan.chen@anu.edu.au) as I did not check github frequently. I will normaly response in a week