keypoints labeling problem

[LiGuo12]

Hi ryo,

We are using your software to complete an annotation on pose estimation. We need to annotate some key points on an image. How can we display multiple key points at the same time? Now, each image only shows one keypoint at a time (the green box in the lower left corner). We are paid users, but we cannot display multiple key points in either rectlabel or rectlabel pro. We need your help.

[ryouchinsa]

Thanks for writing the issue. When you label keypoints, you have 2 options such as Create point and Create keypoints.

It looks that you are using the Create point mode. To show all edit points, check on the "Show all edit points" on the settings.

Create point.

Create keyponts. If you use the Create keyponts mode, you can set the skeleton to connect each keypoint, and trained on popular training tools such as Detectron2 and YOLOv8.

[LiGuo12]

Hi Ryo,

Thanks a lot for this answer, and it is really helpful.

We tried using keypoint mode to do annotation, but the connection (skeleton) between points was not displayed. How can I make the connection between points appear? Will the connection be displayed in the final annotation file(coco, csv, yolo formatn)? If we use point mode to create annotations, do the pose estimation models of detectron2 and yolov8 accept annotations created in point mode? Or are only annotations created in keypoint mode accepted by these pre-trained models?

We previously used rectlabel to label object detection and segmentation, and exported the labels in coco format, which performed very well on the detectron2 pre-trained models. If we use the pretrained pose estimation models of detectron2 and yolov8, should we also export to coco format?

By the way, our team uses rectlabel to annotate our data and cite rectlabel in 3 of our published papers. We really appreciate the software you developed, it works really well.

Best,

Li

On 2024-03-26 00:01, Ryo Kawamura wrote:

Thanks for writing the issue. When you label keypoints, you have 2 options such as Create point and Create keypoints.

It looks that you are using the Create point mode. To show all edit points, check on the "Show all edit points" on the settings.

2024-03-26.15.32.32.png (view on web) [1]

Create point.

2024-03-26.15.32.21.png (view on web) [2]

Create keyponts. If you use the Create keyponts mode, you can set the skeleton to connect each keypoint, and trained on popular training tools such as Detectron2 and YOLOv8.

2024-03-26.15.34.08.png (view on web) [3]

-- Reply to this email directly, view it on GitHub [4], or unsubscribe [5]. You are receiving this because you authored the thread.Message ID: @.***>

Links:

[1] https://github.com/ryouchinsa/Rectlabel-support/assets/1954306/01f92d8c-4054-4be2-9a3e-be86c9404772 [2] https://github.com/ryouchinsa/Rectlabel-support/assets/1954306/94b72db8-3873-4087-8c57-490657680524 [3] https://github.com/ryouchinsa/Rectlabel-support/assets/1954306/3c99ca8e-1df8-4fe4-a35f-83c44e3ea7b6 [4] https://github.com/ryouchinsa/Rectlabel-support/issues/253#issuecomment-2019526579 [5] https://github.com/notifications/unsubscribe-auth/A4PRYFDLPKNL3I36Q6EE3UDY2EMMJAVCNFSM6AAAAABFHZ6I7WVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDAMJZGUZDMNJXHE

[ryouchinsa]

Thanks for your detailed feedback.

To add an edge of the skeleton, drag from one to another point holding option button. When you right click on the point, edit menu opens.

• “Change keypoint name” to change the keypoint name.
• “Change keypoint color” to change the keypoint color and the edge color is defined by the source point color.

Read more about Create keypoints.

If you are using the Create point mode, you have to convert to the keypoints format to train on Detectron2 and YOLOv8. If you could export the current annotations in COCO format and send to us support@rectlabel.com, we can code the conversion script using python.

After conversion, you can import the COCO file to RectLabel, and your annotations will be shown in Create keypoints mode. If you train on Detectron2, please export in COCO format. If you train on YOLOv8, please export in YOLO format.

Thanks for using RectLabel in your team and citing in your papers. If you could provide the url of each paper, it would be very helpful for us.

[LiGuo12]

Hi Ryo,

Thank you for your help.

The title of the published paper is Deep Learning Model Shows Promise for Detecting and Grad-ing Sesamoiditis in Horse Radiographs, L. Guo, X. Yu, A. Thair, A. Rideout, A. Collins, Z. Jane Wang, M. Hore, American Journal of Veterinary Research, 2023 Oct 17, 1(aop), 1-10.DOI: https://doi.org/10.2460/ajvr.23.07.0173

Two other papers are still under review. Thank you for the software, it's very helpful and easy to use.

Best,

Li On 2024-03-27 09:56, Ryo Kawamura wrote:

Thanks for your detailed feedback.

To add an edge of the skeleton, drag from one to another point holding option button. When you right click on the point, edit menu opens.

• “Change keypoint name” to change the keypoint name.
• “Change keypoint color” to change the keypoint color and the edge color is defined by the source point color.

Read more [1] about Create keypoints.

keypoints.gif (view on web) [2]

If you are using the Create point mode, you have to convert to the keypoints format to train on Detectron2 and YOLOv8. If you could export the current annotations in COCO format and send to us @.***, we can code the conversion script using python.

After conversion, you can import the COCO file to RectLabel, and your annotations will be shown in Create keypoints mode. If you train on Detectron2, please export in COCO format. If you train on YOLOv8, please export in YOLO format.

Thanks for using RectLabel in your team and citing in your papers. If you could provide the url of each paper, it would be very helpful for us.

-- Reply to this email directly, view it on GitHub [3], or unsubscribe [4]. You are receiving this because you authored the thread.Message ID: @.***>

Links:

[1] https://rectlabel.com/edit/ [2] https://github.com/ryouchinsa/Rectlabel-support/assets/1954306/f89d1d94-3360-4e0f-8755-0a05a10c68da [3] https://github.com/ryouchinsa/Rectlabel-support/issues/253#issuecomment-2023309012 [4] https://github.com/notifications/unsubscribe-auth/A4PRYFGBA3ZFTO6EMASJBYDY2L225AVCNFSM6AAAAABFHZ6I7WVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDAMRTGMYDSMBRGI

[ryouchinsa]

Thanks for letting us know about your papers and citing RectLabel.

If you need our support for your keypoints annotations, we are happy to support your team.

We updated our help page so that there are gif movies how to use RectLabel including the skeleton. https://rectlabel.com/edit/

Best regards, Ryo

[ryouchinsa]

Currently If you need our support to fix this problem, please let us know.