Update requirements.txt

Update requirements.txt

adding these print statements, we can check if the landmarks are being correctly extracted and if the number of landmarks is as expected for each category

added webcam_interface = gr.Interface( fn=predict_with_webcam, inputs=gr.inputs.Image(shape=(480, 640), source="webcam"), outputs=gr.outputs.Label(), # Use "Label" type instead of "Textbox" live=True, interpretation="default", title="Webcam Landmark Prediction", description="Make predictions using landmarks extracted from your webcam stream.", )

added webcam_interface = gr.Interface( fn=predict_with_webcam, inputs=gr.inputs.Image(shape=(480, 640), source="webcam"), outputs=gr.outputs.Textbox(), # Use "Textbox" type instead of "Label" live=True, interpretation="default", title="Webcam Landmark Prediction", description="Make predictions using landmarks extracted from your webcam stream.", )

modified make prediction and added inv_index_to_class = {v: k for k, v in index_to_class.items()}

added # Modify the Gradio interface to use a "label" type output instead of "textbox" webcam_interface = gr.Interface( fn=predict_with_webcam, inputs=gr.inputs.Image(shape=(480, 640), source="webcam"), outputs=gr.outputs.Label(), # Use "label" type instead of "textbox" live=True, interpretation="default", title="Webcam Landmark Prediction", description="Make predictions using landmarks extracted from your webcam stream.", )

removed share=True

new

move the np.concatenate line inside the extract_keypoints function and make sure to return the concatenated array

rectified indentation

Update app.py

modified # Function to perform holistic detection using Mediapipe def mediapipe_detection(image, model): # COLOR CONVERSION BGR 2 RGB image_rgb = cv2.cvtColor(image, cv2.COLOR_BGR2RGB) image_rgb.flags.writeable = False # Image is no longer writeable results = model.process(image_rgb) # Make prediction image_rgb.flags.writeable = True # Image is now writeable image_bgr = cv2.cvtColor(image_rgb, cv2.COLOR_RGB2BGR) # COLOR CONVERSION RGB 2 BGR return image_bgr, results

defined mp_holistic in predict with webcam fn

using old code

specifying the webcam source explicitly by passing the webcam ID as an argument to the gr.Interface function.

app with exception handling for face landmark

final app before bed

Update app.py

Upload app.py

Upload app.py

Update app.py

moved cap = cv2.VideoCapture(0) to the top

new app.py

fixing return prediction outside function

reformatted code

new make_prediction and predict_with_webcam

removed multiple : cap = cv2.VideoCapture(0)

auto indent, reformat code.

Update app.py

deleted multiple and left one "with mp_holistic.Holistic(min_detection_confidence=0.5, min_tracking_confidence=0.5) as holistic:"

removed the dot

cleared this from 222

Update app.py

put face 1st in the order of landmark extraction

Update app.py

redefined predict_with_webcam . With this modification, the input tensor passed to the PreprocessLayer should have the correct shape, and the error should be resolved.

added N_ROWS = 543 N_DIMS = 3 DIM_NAMES = ['x', 'y', 'z'] SEED = 42 NUM_CLASSES = 250

Update app.py

re-modified code with mediapipe and webcam initialized

Update app.py

Update app.py

Update app.py

Update app.py

adding type numpy to image input

printing processed landmarks

Update app.py

redefined predict_with_webcam function to return "Could not detect landmarks. Make sure your webcam is working properly."

redefined how the webcam is launched

modified predict_with_webcam to continuously predict output.