Unable to Import COCO, PASCAL, YOLO annotations - Error 500 [Urgent Support Requested]

[GitarthVaishnav]

My actions before raising this issue

• Read/searched the docs
• Searched past issues

When I try to import annotations for a small dataset, CVAT online as well as the locally hosted version throws an error 500 with the error

datumaro.components.errors.datasetnotfounderror: failed to find dataset at '/home/django/data/projects/1/tmp/tmptsnzb3r6'\n"..

in the console.

Steps to Reproduce (for bugs)

1. Create a Project "XYZ" with 5 labels.

2. Create a Task and upload images directly (no archive) with local storage option and 100% image quality.

3. Open the job and upload annotations (archive with just annotation files or JSON file) - I have coco json, yolo, and pascal voc xml annotations - which need to be verified and updated.

   • The YOLO data are in the format:
     images
     | 1.jpg
     | 2.jpg
     ...
     labels
     | 1.txt
     | 2.txt
     ...
     data.yaml
     data.names
     obj.data
     train.txt

   • The COCO JSON data is in the same format except that the images folder has a file named _annotations.json, and there are no labels, data.yaml, etc files and folders.

   • The PASCAL VOC XML data is in the same format except that the images folder has 1.xml, 2.xml etc files and there are no labels, data.yaml, etc files and folders.

4. CVAT will take the annotations in all formats, and throw the same error after some time and ask to check console.

5. This is happening with many datasets, also sometimes for 5-7 images it works!

6. Annotations are not flawed - model trains, tested on roboflow, makesense, and labelimg and it works! Want to use CVAT for better processing as always.

Expected Behaviour

• Old behaviour was that when I used CVAT earlier (a couple of months), with the same dataset, it used to take the old yolo format archived labels folder with data.yaml. This does not work.
• Expected behaviour is that with the correct dataset for all formats, it should accept and display the annotations.

Current Behaviour

• error 500

datumaro.components.errors.datasetnotfounderror: failed to find dataset at '/home/django/data/projects/1/tmp/tmptsnzb3r6'\n"..

OR

Possible Solution

• No Idea!

Context

• Annotating data for a project, need urgent support.

Your Environment

• Git hash commit (git log -1):
  commit 2709802 (HEAD -> develop, origin/develop, origin/HEAD)
  Author: Maxim Zhiltsov zhiltsov.max35@gmail.com
  Date: Wed Apr 5 01:59:25 2023 +0300
  Fix cloud storage permissions (Fix cloud storage permissions #5956)

• Docker version docker version (e.g. Docker 17.0.05):

Client:
 Cloud integration: v1.0.31
 Version:           20.10.23
 API version:       1.41
 Go version:        go1.18.10
 Git commit:        7155243
 Built:             Thu Jan 19 17:35:19 2023
 OS/Arch:           darwin/amd64
 Context:           default
 Experimental:      true
Server: Docker Desktop 4.17.0 (99724)
 Engine:
  Version:          20.10.23
  API version:      1.41 (minimum version 1.12)
  Go version:       go1.18.10
  Git commit:       6051f14
  Built:            Thu Jan 19 17:32:04 2023
  OS/Arch:          linux/amd64
  Experimental:     false
 containerd:
  Version:          1.6.18
  GitCommit:        2456e983eb9e37e47538f59ea18f2043c9a73640
 runc:
  Version:          1.1.4
  GitCommit:        v1.1.4-0-g5fd4c4d
 docker-init:
  Version:          0.19.0
  GitCommit:        de40ad0

• Are you using Docker Swarm or Kubernetes?
  Docker
• Operating System and version (e.g. Linux, Windows, MacOS):
  MacOS 13.3 - 2.6 GHz 6-Core Intel Core i7
• Code example or link to GitHub repo or gist to reproduce problem:
  NA
• Other diagnostic information / logs:

[2023-04-05 05:40:54,471] WARNING django.request: Unauthorized: /api/auth/password/change
[2023-04-05 05:40:55,150] WARNING django.request: Not Found: /api/user-agreements
[2023-04-05 05:41:16,658] WARNING django.request: Not Found: /api/functions/requests/
[2023-04-05 05:43:36,643] INFO cvat.server.project_1: label:create Label id:1 for spec:OrderedDict([('name', 'Ready'), ('type', 'any'), ('color', '#689981')]) with sublabels:[], parent_label:None
[2023-04-05 05:43:36,654] INFO cvat.server.project_1: label:create Label id:2 for spec:OrderedDict([('name', 'empty_pod'), ('type', 'any'), ('color', '#a38aca')]) with sublabels:[], parent_label:None
[2023-04-05 05:43:36,666] INFO cvat.server.project_1: label:create Label id:3 for spec:OrderedDict([('name', 'germination'), ('type', 'any'), ('color', '#dff59f')]) with sublabels:[], parent_label:None
[2023-04-05 05:43:36,678] INFO cvat.server.project_1: label:create Label id:4 for spec:OrderedDict([('name', 'pod'), ('type', 'any'), ('color', '#4a8cf4')]) with sublabels:[], parent_label:None
[2023-04-05 05:43:36,690] INFO cvat.server.project_1: label:create Label id:5 for spec:OrderedDict([('name', 'young'), ('type', 'any'), ('color', '#7ea09e')]) with sublabels:[], parent_label:None
[2023-04-05 05:43:37,276] WARNING django.request: Not Found: /api/functions/requests/
[2023-04-05 05:44:48,373] INFO cvat.server: create task #1
[2023-04-05 05:46:17,244] INFO cvat.server: Found frames 1510 for Data #1
[2023-04-05 05:46:17,245] INFO cvat.server: New segment for task #1: idx = 0, start_frame = 0,             stop_frame = 1509
[2023-04-05 05:46:18,390] INFO cvat.server.task_1: get repository request
[2023-04-05 05:46:28,773] WARNING django.request: Not Found: /api/functions/requests/
[2023-04-05 05:46:31,671] INFO cvat.server.task_1: get repository request
[2023-04-05 05:47:26,321] ERROR django.request: Internal Server Error: /api/jobs/1/annotations
[2023-04-05 05:47:51,966] WARNING django.request: Not Found: /api/user-agreements

[zhiltsov-max]

Hi, please make sure the uploaded annotations and datasets use the file layouts described here.

[GitarthVaishnav]

It does work with the exact "word to word" file structure and naming.
Would it need exact naming convention?
obj_train_data, and so on?

Isn't there a generalised file structure?

[zhiltsov-max]

Image file names can be arbitrary, given they are listed in the corresponding subset file (train.txt, valid.txt). Everything else should be the same as in the description. This file layout was borrowed from the YOLO training framework, meaning it should be ready-to-use in the practical use cases. Could you please specify which elements you would like to be optional in the file structure?

[elhabchiali]

If anyone has the same problem, i was able to solve it for the YOLO case by including the images alongside the annotation files.

[rafaelgildin]

@elhabchiali did you export the images with yolo format from roboflow ? If not any advice on how to export roboflow dataset to cvat ? Thanks.