app.py

import cv2
import numpy as np
import gradio as gr
from tensorflow.keras.models import load_model

# Define constants
SEQUENCE_LENGTH = 20  # Number of frames to extract
IMAGE_HEIGHT = 64  # Height of each frame
IMAGE_WIDTH = 64  # Width of each frame
CLASSES_LIST = ["Archery", "BabyCrawling", "Balance_Beam", "EyeMakeup", "LipStick"]

# Load the model
loaded_model = load_model(r"LRCN_model.h5")

def frames_extraction(video_reader):
    frames_list = []
    video_frames_count = int(video_reader.get(cv2.CAP_PROP_FRAME_COUNT))
    skip_frames_window = max(int(video_frames_count / SEQUENCE_LENGTH), 1)  # default skip is 1
    for frame_counter in range(SEQUENCE_LENGTH):
        video_reader.set(cv2.CAP_PROP_POS_FRAMES, frame_counter * skip_frames_window)
        success, frame = video_reader.read()
        if not success:
            break

        resized_frame = cv2.resize(frame, (IMAGE_HEIGHT, IMAGE_WIDTH))
        normalized_frame = resized_frame / 255  # color 0-255
        frames_list.append(normalized_frame)
    video_reader.release()

    return frames_list

# Function to classify the video
def classify_video(frames):
    predicted_labels = np.argmax(loaded_model.predict(frames), axis=1)
    predicted_class_label = CLASSES_LIST[predicted_labels[0]]  # Ensure we get the label for the first prediction
    return predicted_class_label

# Define the prediction function
def predict_video(video_file):
    video_capture = cv2.VideoCapture(video_file)
    features = []
    video_reader = video_capture
    frames = frames_extraction(video_reader)
    if len(frames) == SEQUENCE_LENGTH:
        features.append(frames)
        features = np.asarray(features)
        predicted_class = classify_video(features)
        video_capture.release()
        return predicted_class

# Gradio interface definition
iface = gr.Interface(
  fn=predict_video,
  inputs=gr.Video(),
  outputs="text",
  title="Action Recognition with LSTM",
  description="Upload a video and get the predicted action class."
)

# Launch the Gradio interface
iface.launch()