Stuck on an issue?

Lightrun Answers was designed to reduce the constant googling that comes with debugging 3rd party libraries. It collects links to all the places you might be looking at while hunting down a tough bug.

And, if you’re still stuck at the end, we’re happy to hop on a call to see how we can help out.

Is it possible to train a MOT model on one class of LaSOT dataset?

See original GitHub issue

Hello, I want to make a model that can track multiple drones in a video using only drone data from LaSOT dataset. I’d appreciate it if you let me know if this is a possible idea in the first place.

Preliminaries

I downloaded only the drone folder from the LaSOT dataset and made it into CocoVideoDataset.
The data format is as follows.

data
ㄴ lasot
    ㄴ annotations
        ㄴ lasot_train.json
        ㄴ lasot_test.json
        ㄴ lasot_train_infos.txt
        ㄴ lasot_test_infos.txt
    ㄴ LaSOTBenchmark
        ㄴ drone
            ㄴ drone-1
            ㄴ drone-2
            ㄴ ....

Created a custom config in configs/mot/bytebrack. base also uses two custom modules '../../_base_/datasets/lasot_drone.py' '../../_base_/models/custom_yolox_x_8x8.py'.
Changed to num_classes=1, classes=(‘drone’, ) in all configuration files.

===

Error

When training starts, after normal training for only 1 epoch, data['category_id'] = self.cat_ids[label] error occurs in the evaluation process. I’ve looked all over for similar issues in mmDetection, but they didn’t solve it. Do you know any solution?

Issue Analytics

State:
Created a year ago
Comments:13

Top GitHub Comments

1reaction

pixeli99commented, May 27, 2022

Of course, you can take a look at the issue🚀.

https://github.com/open-mmlab/mmtracking/issues/564

0reactions

moey920commented, Aug 1, 2022

It works good. However, I’m exploring the architecture of the detector because the class I’m trying to detect has poor performance. Try it.

Top Results From Across the Web

Training sequences are generated from the public MOT ...

Training sequences are generated from the public MOT dataset. Each training sequence has detections from the same object throughout the track and one...

LaSOT — Large-Scale Dataset for Object Tracking Models

LaSOT is a large-scale dataset for training and evaluating object tracking models. The dataset contains 1.5 thousand video clips with objects of 85...

Object Tracking and Reidentification with FairMOT

In this blog post, we will focus on one such tracker, FairMOT, ... The input image is passed to an object detection model....

MOT17 Dataset - Papers With Code

The Multiple Object Tracking 17 (MOT17) dataset is a dataset for multiple ... A video for each scene is divided into two clips,...

mmtrack.apis — MMTracking 0.14.0 documentation

Inference image(s) with the mot model. ... mmtrack.apis.train_model(model, dataset, cfg, distributed=False, ... LaSOT dataset of single object tracking.