Added yolov7 into the sahi model.py #1
Conversation
kimb3rlyn
changed the title
|
@kimb3rlyn change the camel case to snake case leh. Class names are still okay |
scripts/inferenceSahi.py
Outdated
| model = Yolov7DetectionModel( | ||
| image_size = 1280, | ||
| model_path = weightPath, | ||
| config_path = cfgPath | ||
| ) | ||
|
|
There was a problem hiding this comment.
expose some more parameters here so people know what are the main things to change
There was a problem hiding this comment.
These are the main parameters to change. The rest are optional on the model
| category_remapping: Optional[Dict] = None, | ||
| load_at_init: bool = True, | ||
| image_size: int = None, | ||
| bgr: bool = False, |
There was a problem hiding this comment.
is there a point to expose this then if sahi uses PIL? will there be a situation where we need this to be true?
There was a problem hiding this comment.
For future development maybe; where SAHI no longer user PIL ? cause is still under development
|
Add sahi into requirements.txt, maybe update readme on how to use too |
Completed |
|
Found a potential bug in yolov7/yolov7/utils/sahiModelExtended.py Line 287 in fafbf8f currently it is using the prediction results from detect_get_box_in(). However, this function already has NMS applied to results before the merging. Think it should be something more similar to https://github.com/obss/sahi/blob/33e215019519869936a997718bcba404a63d28f6/sahi/models/yolov5.py#L53 instead, so we can perform NMS after merging the results |
| except Exception as e: | ||
| raise TypeError("Error loading model yolov7 : ", e) | ||
|
|
||
| def perform_inference(self, image: np.ndarray): |
|
bump! @yhsmiley @kimb3rlyn would like to use this hehehe |
Added yolov7 into SAHI model
Need to pip install SAHI library first before using