Adding palm_detection description

tensor-groups/palm-detection-out.md

@@ -0,0 +1,66 @@
+# Classification
+
+- tensor-group: yes
+- layer-type: output
+- use-case: hand-detection
+
+# Description
+
+Palm Detection model outputs. This tensor group contains detection results for identifying palm/hand regions in images. The Palm Detection model is designed to locate hand bounding boxes and keypoints within an image.
+
+Palm Detection Output Tensors:
+
+| Name                | shape    |
+|---                  |---       |
+| [detections]        | N x 8    |
+
+Where N is the number of detected palms/hands.
+
+# Tensor Decoding Logic
+
+```
+Foreach detection in detections:
+    score = detection[0]
+    box_x = detection[1]
+    box_y = detection[2]
+    box_size = detection[3]
+    kp0_x = detection[4]
+    kp0_y = detection[5]
+    kp2_x = detection[6]
+    kp2_y = detection[7]
+
+    # Filter by confidence threshold (typically 0.60)
+    if score > confidence_threshold:
+        output_detection = [score, box_x, box_y, box_size, kp0_x, kp0_y, kp2_x, kp2_y]
+```
+
+## Calculate the orientation of the bounding box
+
+```
+d_x = kp2_x - kp0_x
+d_y = kp2_y - kp0_y
+
+angle = atan2 (-d_y, d_x )
+```
+**d_y** is negated because, in image coordinate systems, the y-axis typically increases downward, opposite to the conventional mathematical coordinate system.
+**angle** is measured in radians.
+
+# External References
+
+* [MediaPipe Hand Detection](https://google.github.io/mediapipe/solutions/hands.html)
+* [Palm Detection Model Architecture](https://github.com/google/mediapipe/blob/master/mediapipe/modules/hand_landmark/hand_detection.pbtxt)
+* [Hand Gesture Recognition Using ONNX](https://github.com/PINTO0309/hand-gesture-recognition-using-onnx)
+
+# Models
+
+* [palm_detection_full_inf_post_192x192.onnx]
+
+# Tensor Decoders
+
+|Framework | Links |
+|---       |---    |
+|ONNX / Python | [Hand gesture recognition using ONNX](https://github.com/PINTO0309/hand-gesture-recognition-using-onnx) |
+
+
+[detections]: /tensors/palm-detection-out-detections.md
+[palm_detection_full_inf_post_192x192.onnx]: /models/palm_detection_full_inf_post_192x192.onnx

tensor-id-register.md

@@ -13,6 +13,7 @@
 |yolo-v8-segmentation-out-detections | [details](/tensors/yolo-v8-segmentation-out-detections.md) |
 |yolo-v8-segmentation-out-protos | [details](/tensors/yolo-v8-segmentation-out-protos.md) |
 |yolo-v8-out | [details](/tensors/yolo-v8-out.md) |
+|palm-detection-out | [details](/tensors/palm-detection-out-detections.md) |
 
 # Tensor Groups (Model Family)
 
@@ -21,3 +22,4 @@
 |ssd-mobilenet-v1-variant-1-out | [details](/tensor-groups/ssd-mobilenet-v1-variant-1-out.md) |
 |ultra-lightweight-face-detection-rfb-320-v1-variant-1-out | [details](/tensor-groups/ultra-lightweight-face-detection-rfb-320-v1-variant-1-out.md) |
 |yolo-v8-segmentation-out | [details](/tensor-groups/yolo-v8-segmentation-out.md) |
+|palm-detection-out | [details](/tensor-groups/palm-detection-out.md) |

tensors/palm-detection-out-detections.md

@@ -0,0 +1,71 @@
+# Classification
+
+- tensor-group: no
+- layer-type: output
+- use-case: hand-detection
+- part-of-tensor-groups:
+    - [palm-detection-out](/tensor-groups/palm-detection-out.md)
+
+# Description
+
+Palm detection predictions containing bounding box location, confidence score, and hand keypoints from the Palm Detection model. This tensor contains detection results for identifying palm/hand regions in images.
+
+## Detections Tensor
+
+- tensor-shape: N x 8
+- tensor-datatype: float32
+
+Where:
+ * N is the number of detected palms/hands
+
+### Known Aliases
+* pdscore_boxx_boxy_boxsize_kp0x_kp0y_kp2x_kp2y
+
+### Encoding
+
+Scheme: (Score, Box_X, Box_Y, Box_Size, Keypoint0_X, Keypoint0_Y, Keypoint1_X, Keypoint1_Y)
+
+The tensor contains 8 channels with the following layout per detection:
+
+|Channel 0  | Channel 1 | Channel 2 | Channel 3  | Channel 4 | Channel 5 | Channel 6 | Channel 7 |
+|---        |---        |---        |---         |---        | ---       |---        | ---  |
+| Score     | Box_X     | Box_Y     | Box_Size   | Keypoint0-x   | Keypoint0-y | Keypoint1-x | Keypoint1-y   |
+
+
+Memory layout of tensor data (for detection i):
+
+|Index           | Symbol            | Value              | Comment                                    |
+|---             |---                |---                 |---                                         |
+| -              | -                 | -                  | -                                          |
+|0               | pd_score          | detection-i-confidence | tensor-start, detection-i-score            |
+|1               | box_x             | detection-i-box-center-x | tensor-continue, detection-i-box-x         |
+|2               | box_y             | detection-i-box-center-y | tensor-continue, detection-i-box-y         |
+|3               | box_size          | detection-i-size   | tensor-continue, detection-i-size          |
+|4               | keypoint0_x             | detection-i-keypoint0-x | tensor-continue, detection-i-keypoint0-x   |
+|5               | keypoint0_y             | detection-i-keypoint0-y | tensor-continue, detection-i-keypoint0-y   |
+|6               | keypoint1_x             | detection-i-keypoint2-x | tensor-continue, detection-i-keypoint1-x   |
+|7               | keypoinp1_y             | detection-i-keypoint2-y | tensor-end, detection-i-keypoint1-y        |
+
+### Details
+
+- **pd_score** (Channel 0): Confidence score for the detection, in range [0, 1]
+- **box_x, box_y** (Channels 1-2): Center coordinates of the detected palm bounding box in range [0,1] where 0 corresponds to the top (or left) and 1 to bottom (or right) edge of the image
+- **box_size** (Channel 3): Side length of the bounding box, where the bounding box is square
+- **Keypoint0** (Channels 4-5): Wrist keypoint 
+- **Keypoint1** (Channels 6-7): Middle finger joint to palm
+
+
+# External References
+
+* [MediaPipe Hand Detection](https://google.github.io/mediapipe/solutions/hands.html)
+* [Palm Detection Model Architecture](https://github.com/google/mediapipe/blob/master/mediapipe/modules/hand_landmark/hand_detection.pbtxt)
+
+# Models
+
+* [palm_detection_full_inf_post_192x192.onnx](https://gitlab.collabora.com/gstreamer/onnx-models/-/blob/master/models/palm_detection_full_inf_post_192x192.onnx)
+
+# Tensor Decoders
+
+|Framework | Links |
+|---       |---    |
+|ONNX    | [Hand gesture recognition using ONNX](https://github.com/PINTO0309/hand-gesture-recognition-using-onnx/blob/main/model/palm_detection/palm_detection.py) |