Suport ultralytics yolo pose and seg model. Optimized fuzzy matching for unsupported models.

[315386775]

Update: Added fuzzy matching for unsupported models.

Scenario: When I enter a non-existent model ID, I will be prompted that the model is not supported while also being given suggestions for similar models. This is especially relevant when dealing with the YOLO series of models. For example, if the original YOLO model ID supports "ultralytics/yolov11s-seg" and also supports "yolov11s-seg", but I input something that doesn't exactly match, I should receive a suggestion to use one of the supported model IDs.

Compare result:

model = xinfer.create_model("ultralytics/yolo11n-pose", device="cuda", dtype="float16")

# fuzzy matching results
ValueError: Unsupported model: ultralytics/yolo11n-pose. Suggestion model: ultralytics/yolov11n-pose, ultralytics/yolov11n-seg, ultralytics/yolov11s-seg。

# original results
ValueError: Unsupported model: ultralytics/yolo11n-pose

Update: Suport ultralytics yolo pose and seg model

Pose model output:

model = xinfer.create_model("ultralytics/yolov11n-pose", device="cuda", dtype="float16")

[[[{'point': [633.5, 442.0], 'score': 0.9853515625, 'name': 'Nose'}, {'point': [667.0, 378.5], 'score': 0.814453125, 'name': 'Left Eye'}, {'point': [554.5, 364.25], 'score': 0.998046875, 'name': 'Right Eye'}, {'point': [0.0, 0.0], 'score': 0.0054473876953125, 'name': 'Left Ear'}, {'point': [360.5, 382.25], 'score': 0.99609375, 'name': 'Right Ear'}, {'point': [663.0, 669.0], 'score': 0.669921875, 'name': 'Left Shoulder'}, {'point': [198.5, 676.0], 'score': 0.52880859375, 'name': 'Right Shoulder'}, {'point': [0.0, 0.0], 'score': 0.01554107666015625, 'name': 'Left Elbow'}, {'point': [0.0, 0.0], 'score': 0.00963592529296875, 'name': 'Right Elbow'}, {'point': [0.0, 0.0], 'score': 0.033477783203125, 'name': 'Left Wrist'}, {'point': [0.0, 0.0], 'score': 0.0341796875, 'name': 'Right Wrist'}, {'point': [0.0, 0.0], 'score': 0.006439208984375, 'name': 'Left Hip'}, {'point': [0.0, 0.0], 'score': 0.003635406494140625, 'name': 'Right Hip'}, {'point': [0.0, 0.0], 'score': 0.004573822021484375, 'name': 'Left Knee'}, {'point': [0.0, 0.0], 'score': 0.002368927001953125, 'name': 'Right Knee'}, {'point': [0.0, 0.0], 'score': 0.004756927490234375, 'name': 'Left Ankle'}, {'point': [0.0, 0.0], 'score': 0.0021076202392578125, 'name': 'Right Ankle'}]]]
Saved Render Imgae to res/1.jpg

Seg Model output:

model = xinfer.create_model("ultralytics/yolov11n-seg", device="cuda", dtype="float16")

[[{'mask': array([[          0,           0,           0, ...,           0,           0,           0],
       [          0,           0,           0, ...,           0,           0,           0],
       [          0,           0,           0, ...,           0,           0,           0],
       ...,
       [          0,           0,           0, ...,           0,           0,           0],
       [          0,           0,           0, ...,           0,           0,           0],
       [          0,           0,           0, ...,           0,           0,           0]], dtype=float32), 'bbox': [87.0, 101.0, 765.0, 713.0], 'category_id': 0, 'score': 0.5380859375, 'class_name': 'person'}]]
Saved Render Imgae to res/1.jpg