You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
These models are quite old and limited. For example, cocossd supports only 90 labels, which is too small. Included yolo works much worse and also has the same 90 labels.
there is a support for URL of a compatible model instead of the model name. I think it should work like this
Could someone share any URLs, which will work with ML5JS object detector version 0.12.2?
I did find some models, like this (YOLOv8) https://github.com/ultralytics/ultralytics/tree/v8.2.94
it supports 600 labels, but it only works for 640x640 pixels images, while the included cocossd model works for any image size.
Dear ml5 community,
I'm submitting a new request for information/advice. Please see the details below.
β Step 1: Describe the issue π
When I use objectDetector in v0.12.2 it supports 2 models: cocossd and yolo as mentioned here: https://archive-docs.ml5js.org/#/reference/object-detector
These models are quite old and limited. For example, cocossd supports only 90 labels, which is too small. Included yolo works much worse and also has the same 90 labels.
there is a support for URL of a compatible model instead of the model name. I think it should work like this
detector = ml5.objectDetector('https://URL/to/model.json', {modelUrl: 'https://URL/to/model.json'}, modelReadyCallbackFunc);
but I could not find any compatible models.
Could someone share any URLs, which will work with ML5JS object detector version 0.12.2?
I did find some models, like this (YOLOv8)
https://github.com/ultralytics/ultralytics/tree/v8.2.94
it supports 600 labels, but it only works for 640x640 pixels images, while the included cocossd model works for any image size.
Also, there are some models like Darknet
https://cdn.jsdelivr.net/gh/ml5js/ml5-data-and-models@master/models/darknetclassifier/darknetreference/model.json
but they do not work. I guess it is because they were created for image classification, not object detection.
The text was updated successfully, but these errors were encountered: