DEEPCRAFT™ Ready Models
The table below lists the Ready Models available as DEEPCRAFT™ Studio projects. Refer to the Ready Model Reports for performance metrics, test results, and on-device measurements.
| Model | Modality | Typical use |
|---|---|---|
| Baby Cry Detection | Audio | Wearables and baby products; children 0–4 years |
| Cough Detection | Audio | Health wearables and nearby devices |
| Siren Detection | Audio | Wearables, hearables, in-cabin vehicle systems |
| Snore Detection | Audio | Sleep wearables and near-bed devices |
| Gesture Classification | Radar | Touchless control of a monitor or connected device |
Baby Cry Detection, Cough Detection, Siren Detection, and Snore Detection each detect one sound event in mono audio sampled at 16 kHz. Gesture Classification detects five hand gestures from an Infineon XENSIV™ 60 GHz radar sensor.
All recommendations, performance information, and other details refer exclusively to the Ready Model included in the project. Any modification to the model — such as retraining, altering its architecture, or changing its parameters — is likely to invalidate these details, including the stated recommendations and performance metrics.
DEEPCRAFT™ Ready Model for Baby Cry Detection
The DEEPCRAFT™ Ready Model for Baby Cry Detection is designed to detect crying in babies and young children between 0 and 4 years old. This model can be used in a wearable device or smart baby product to alert the parents of a crying or active baby.
Operating recommendations
- Detection range: The Ready Model performs optimally within a range of up to 5 meters. Beyond this distance, performance may degrade, and cry detection accuracy can decrease.
- Confidence threshold: The recommended prediction confidence level is 0.85, with an acceptable range of [0.7, 0.85].
- Post-processing: Detection is confirmed only if 3 consecutive predictions exceed the confidence threshold.
Refer to the Baby Cry Detection Ready Model Report for performance metrics and test results.
DEEPCRAFT™ Ready Model for Cough Detection
The DEEPCRAFT™ Ready Model for Cough Detection is designed to detect coughs from adults with the purpose of identifying users’ health degradation. The model is intended to run on a wearable device such as a smart watch, bracelet, armband, or necklace, or on a smartphone or other non-wearable device located in the vicinity of the person.
Operating recommendations
- Detection range: The Ready Model performs optimally within a range of up to 2 meters. Beyond this distance, performance may degrade, and cough detection accuracy can decrease.
- Confidence threshold: The recommended prediction confidence level is 0.8, with an acceptable range of [0.7, 0.85].
- Post-processing: No additional post-processing is required.
Refer to the Cough Detection Ready Model Report for performance metrics and test results.
DEEPCRAFT™ Ready Model for Snore Detection
The DEEPCRAFT™ Ready Model for Snore Detection is designed to identify snoring events when the user is sleeping. This model is suitable for use in wearables equipped with a built-in microphone. It is capable of detecting snores in various typical sleeping environments.
Operating recommendations
- Detection range: The Ready Model performs optimally within a range of up to 2 meters. Beyond this distance, performance may degrade, and snore detection accuracy can decrease.
- Confidence threshold: The recommended prediction confidence level is 0.7, with an acceptable range of [0.65, 0.8].
- Post-processing: No additional post-processing is required.
Refer to the Snore Detection Ready Model Report for performance metrics and test results.
DEEPCRAFT™ Ready Model for Gesture Classification
The DEEPCRAFT™ Ready Model for Gesture Classification detects five hand gestures — Push, Swipe Up, Swipe Down, Swipe Left, and Swipe Right — performed in front of an Infineon XENSIV™ 60 GHz radar sensor. This model is suitable for touchless control of a monitor or any other device connected to the radar sensor.
Operating recommendations
- Detection range: The Ready Model performs optimally when the gesture is executed between 10 cm and 70 cm from the radar sensor, within a 10° angle of its field of view. Beyond this range, performance may degrade, and gesture detection accuracy can decrease.
- Confidence threshold: The recommended prediction confidence level is 0.90.
- Post-processing: The Ready Model includes a debounce step, available as the GestureReadyModelPostprocessor Graph UX unit. A gesture starts accumulating detections once its confidence exceeds 0.2, and is confirmed once it reaches the recommended confidence threshold or has been the best-scoring class for 3 consecutive frames. Its counters reset after 4 consecutive low-confidence frames, and a confirmed gesture must be followed by at least 1 low-confidence frame before it can be confirmed again.