dusty-nv / jetson-inference

Hello AI World guide to deploying deep-learning inference networks and deep vision primitives with TensorRT and NVIDIA Jetson.
https://developer.nvidia.com/embedded/twodaystoademo
MIT License
7.75k stars 2.97k forks source link

Inquiry about Retraining Segmentation Model and Formatting VOC Dataset #1803

Closed untimatrix closed 6 months ago

untimatrix commented 6 months ago

Dear Dusty,

I am reaching out to seek guidance on a couple of tasks related to working with segmentation models and custom datasets.

Firstly, I am interested in retraining a segmentation model for a custom VOC dataset. Specifically, I would like to know the steps involved in retraining a segmentation model, starting from a pre-trained model, using a custom VOC dataset. Any insights or resources you could provide on this topic would be greatly appreciated.

Secondly, I have encountered a challenge in formatting my VOC dataset generated by RoboFlow. Currently, my dataset is organized into separate train, test, and validation folders each folder contains both jpg images and xml annotation files, and I need to reformat it to meet the requirements of the object detection SSD model. Could you please advise on the process of restructuring the dataset to adhere to the desired format, particularly with regard to creating the default.txt file under the ImageSets/Main directory? If manual formatting is necessary, I would like to know the contents of the default.txt file for proper organization.

Thank you very much for your time and assistance.

Best regards,