The list of tests presented here is from Benchmark version 1.0 running on the YOLOvX App. The models selection is to make sure that all the test are able to run on almost all the mobile devices, even on low resource devices, so that we get to know how they perform.
Detection
Test 1: YOLOv3-tiny-320
COCO mAP
One of the earliest YOLO-based object detection to breakthrough state-of-the-art and still continues to be popular particularly for edge and mobile deployments.
Detection
Test 2: YOLOv9-tiny-320
Test 3: YOLOv9-tiny-640
COCO mAP
One of the efficient YOLO-based object detection for learning what you want to learn Using Programmable Gradient Information, and is one of the most popular for edge and mobile deployments, here it is YOLOv9 tiny model.
Detection
Test 4: YOLOv9-small-320
COCO mAP
YOLOv9 small variant is supposedly little bigger than tiny model and also achieving higher accuracy.
Detection
Test 5: FaceModel-320
Face16 mAP
Custom trained YOLOv9-based object detection to detect faces, we curated some famous faces for the dataset.
Segmentation
Test 6: YOLOv9-tiny-320
COCO mAP
YOLOv9 model for object segmentation and most popular particularly for edge and mobile deployments.
Segmentation
Test 7: YOLOv9-tiny-640
COCO mAP
YOLOv9 tiny segmentation model with higher input image resolution.
Segmentation
Test 8: CrackDamageModel-320
Test 9: CrackDamageModel-640
COCO mAP
Custom trained YOLOv9-based model for crack & damage detection on walls. The images for dataset are curated for test benchmark.