hand_gesture_kp: Verified on the ECS-DoT EVB
Small fully connected network that classifies a hand gesture (open hand, closed fist, pointing) from the 21 hand keypoints of a hand-landmark detector such as MediaPipe Hands.
Compiled with ECS-Compiler in ECS-Studio 1.140.0 and verified on a physical ECS-DoT EVB.
Getting Started
This repository is a ready-to-use model folder for ECS-Compiler, the AI pipeline of ECS-Studio. It holds the ONNX model and its calibration and evaluation data. The verified result was built with the default ECS-Compiler settings; keep them unchanged to reproduce it.
Compile the model
Download the model folder, either from the Files tab of this page or with the Hugging Face CLI:
hf download emass-ai/hand_gesture_kp --local-dir hand_gesture_kpIn ECS-Studio (1.140.0 or later), open a project for the ECS-DoT EVB, open the ECS-Compiler panel and choose
keypoint_classifier.onnxas the model file. The panel loadskeypoint_classifier_train.npz(calibration) andkeypoint_classifier_test.npz(evaluation) from the same folder automatically.Click Start Pipeline, then Build and Upload to run the model on the board. The ECS-Compiler User Guide describes each step.
Model Details
| Model type | gesture_classification |
| Architecture | MLP: Linear 42β20 β ReLU β Linear 20β10 β ReLU β Linear 10β3 β Softmax |
| Input | input_1 [1, 42] float32 (21 hand keypoints, (x, y) each) |
| Preprocessing | Keypoint coordinates relative to the wrist (keypoint 0), flattened, divided by the largest absolute value so they lie in β1 β¦ 1 |
| Output | Identity [1, 3] float32 |
| Labels | Open, Close, Pointer |
| Model size (float) | 5.6 KB |
| Parameters | 1,103 |
Performance Summary
| Model | Precision | Board | Inference Time | Library Size | Compute Unit |
|---|---|---|---|---|---|
| hand_gesture_kp | int8 | ECS-DoT EVB | 711 Β΅s | 25.6 KB | DLA + CPU |
Inference time is the measured time of one dla_infer() call on the board (median of 204 runs, ECS-DoT at 50 MHz). Library size is the compiled model library libdla.a (5 layers on the DLA, 1 on the CPU).
On-board verification: Run on 3 real samples. The board's output matched the compiler's expected output exactly, and its prediction matched ONNX Runtime on every sample (cosine similarity 0.9999 or higher). On the 200-sample evaluation set the board (int8) scored 99.50% against 99.50% for the float model on the same samples, and agreed with ONNX Runtime on 200 of 200 samples.
Accuracy
96.63% top-1 accuracy on Hand gesture keypoint dataset of the hand-gesture-recognition-using-mediapipe project, held-out test split (889 samples, Apache-2.0), measured with ONNX Runtime on the float model.
Training Data
- Hand gesture keypoint dataset of the hand-gesture-recognition-using-mediapipe project: https://github.com/kinivi/hand-gesture-recognition-mediapipe (licence: Apache-2.0).
- Citation: Kazuhito Takahashi, Nikita Kiselov, hand-gesture-recognition-using-mediapipe (Apache-2.0).
Files
| File | Contents | Download |
|---|---|---|
keypoint_classifier.onnx |
The ONNX model | Download |
keypoint_classifier_train.npz |
Calibration set the board build quantizes with | Download |
keypoint_classifier_test.npz |
Evaluation set | Download |
sample_input.npz |
The real samples run on the board (x, with labels y) |
Download |
LICENSE |
License | Download |
README.md |
This model card | Download |
Intended Use
- This model is provided to evaluate and validate ECS-Compiler and the ECS-DoT. It is not a product and has not been tested for any particular end use.
License
- This model is released under
apache-2.0; seeLICENSE. The training data keeps its own licence (see Training Data).
Provenance
- Verified with ECS-Studio 1.140.0 (ECS-Compiler engine 0.6.0) on the ECS-DoT EVB, 2026-11-17. Build fingerprint
3ead310b4215.