Integrate emotions recognition #51
Conversation
integraledelebesgue
changed the title
| face_emotions = [] | ||
| boxes = [] | ||
| frame_height, frame_width, _ = frame.shape | ||
| for face_box in faces.boxes: |
There was a problem hiding this comment.
faces.boxes reference is not safe since faces can be None. This function should return Result | None and perform a check on faces
There was a problem hiding this comment.
Btw, predict method should take not-none arguments. You can rename this method to __predict_safe and use it in stream and predict_batch.
| emotions: list[float] | ||
| boxes: list[FloatArray2] |
There was a problem hiding this comment.
We could stack these results and store them as np.ndarrays instead of lists. We do so in other Results in general
There was a problem hiding this comment.
Also, it's no use duplicating face.Result as boxes here
There was a problem hiding this comment.
There is - that way there is no coupling between emotion recognition and face detection in the visualization layer.
Duplicating it is almost free (the cost of doing that is negligible), but it's much easier to handle it that way in the visualization layer.
| class Result: | ||
| emotions: list[float] | ||
| boxes: list[FloatArray2] | ||
|
|
||
| def __init__(self, emotions: list[float], boxes: list[FloatArray2]) -> None: | ||
| self.emotions = emotions | ||
| self.boxes = boxes |
There was a problem hiding this comment.
This should be a frozen dataclass
| scores = { | ||
| 'angry': -0.05, | ||
| 'disgust': 0, | ||
| 'fear': -0.07, | ||
| 'happy': 1, | ||
| 'sad': -1, | ||
| 'surprise': 0, | ||
| 'neutral': 0, | ||
| } |
There was a problem hiding this comment.
This dict shouldn't be located here since it is going to be created on each call 💀 Please make it a private constant
There was a problem hiding this comment.
With this in mind - how about making this dict injected into an Estimator via constructor?
| emotions: list[float] | ||
| boxes: list[FloatArray2] |
There was a problem hiding this comment.
Also, it's no use duplicating face.Result as boxes here
WojciechBarczynski
changed the title
Closes #17
Changes
task.emotion.Estimatorallows estimating emotional "score" of an actor based on the face detectionvisualization.Visualizerannotates face bounding boxes with emotion labels