Demo

demo.mp4

models used:

• bbox detector for finding clock face in the image
• classifier for clock orientation estimation
• keypoint detection for center and top
• semantic segmentation for finding clock hands
• KDE for splitting the binary segmentation mask into individual clock hands

Watch crop with center and top keypoint

Detected mask of watch hands

KDE of pixel angles

Fitted lines to segmented pixels

Final selected and rejected lines

Metrics

Path	val.1-min_acc	val.10-min_acc	val.60-min_acc
metrics/end_2_end_summary.json	0.224	0.345	0.414

Path

Path

Path	eval.iou_score	eval.loss	step	train.iou_score	train.loss
metrics/segmentation.json	0.585	0.262	149	0.851	0.081

Graph

flowchart TD
	node1["datasets/watch-faces.json.dvc"]
	node2["download-images"]
	node3["eval-detector"]
	node4["eval-end-2-end"]
	node5["eval-keypoint"]
	node6["eval-segmentation"]
	node7["export-detector"]
	node8["generate-detection-dataset"]
	node9["generate-watch-hands-dataset"]
	node10["train-detector"]
	node11["train-keypoint"]
	node12["train-segmentation"]
	node13["update-metrics"]
	node1-->node2
	node2-->node3
	node2-->node4
	node2-->node8
	node2-->node9
	node2-->node11
	node2-->node12
	node3-->node13
	node4-->node13
	node5-->node13
	node7-->node3
	node8-->node7
	node8-->node10
	node10-->node4
	node10-->node5
	node10-->node6
	node10-->node7
	node11-->node4
	node11-->node5
	node11-->node13
	node12-->node4
	node12-->node6
	node12-->node13
	node14["example_data/IMG_1200_720p.mov.dvc"]
	node15["render-demo"]
	node14-->node15
	node16["checkpoints/segmentation.dvc"]
	node17["checkpoints/detector.dvc"]
	node18["checkpoints/keypoint.dvc"]

Installation

Install watch_recognition module, run pip in the main repository dir

pip install watch_recognition/

Tested on Python 3.7 and 3.8

Running models

Checkout example notebook: notebooks/demo-on-examples.ipynb

Models description

TODO