count_max_bird_species_tool.py

import os
from sample_youtube_video import sample_youtube_video
from ultralytics import YOLO
from langchain_core.tools import tool
from utils import decode_base64_to_frame

@tool
def count_max_bird_species_in_video(youtube_url: str) -> int:
    """
    Count the maximum number of bird species in a YouTube video.

    Args:
    - youtube_url: str: the URL of the YouTube video

    Returns:
    - int: The maximum number of bird species detected in the YouTube video.
    """

    # frames = sample_youtube_video(youtube_url, 5)
    BASE_DIR = os.path.dirname(os.path.abspath(__file__))
    youtube_video_in_base_64 = os.path.join(BASE_DIR, 'files', 'frames_youtube_birds_video.b64')
    print('youtube video path: ' + youtube_video_in_base_64)
    frames = read_list_from_txt(youtube_video_in_base_64)
    
    return count_max_bird_species(frames)

def count_max_bird_species(framesStr: list[str]) -> int:
    """
    Count the maximum number of bird species in a single frame considering a list of frames.

    Args:
    - frames (List[cv2.Mat]): A list of frames (images) to analyze.

    Returns:
    - int: The maximum number of bird species detected in any single frame.
    """
    
    # List of bird class indices from COCO dataset (class IDs for birds)
    bird_class_ids = [
        14, 15, 16, 17, 18, 19, 20, 21  # Example bird class IDs in COCO dataset
    ]

    # Load YOLOv8 model for object detection (pre-trained or custom-trained for animals)
    model = YOLO("yolov8n.pt")  # Use a pre-trained YOLOv8 model

    max_bird_species_count = 0  # To track the maximum count of distinct bird species

    frames = [decode_base64_to_frame(b64) for b64 in framesStr]

    # Iterate through all frames in the list
    for frame in frames:
        # Perform inference on the current frame
        results = model(frame)

        bird_species_detected = set()  # To track distinct bird species in this frame

        # Parse the detected objects
        for result in results:
            # Convert the class indices from tensor to integer
            detected_classes = result.boxes.cls.int()  # Convert tensor to int

            # Iterate over detected classes and filter only bird species
            for detected_class in detected_classes:
                # If the class is a bird (based on COCO bird class IDs)
                if detected_class.item() in bird_class_ids:
                    species_name = result.names[detected_class.item()]  # Get the species name
                    bird_species_detected.add(species_name)

        # Update the maximum number of distinct bird species detected in any frame
        max_bird_species_count = max(max_bird_species_count, len(bird_species_detected))

    # Return the maximum number of distinct bird species detected in any frame
    return max_bird_species_count

def read_list_from_txt(file_path: str) -> list[str]:
    with open(file_path, 'r', encoding='utf-8') as f:
        return [line.rstrip('\n') for line in f]

if __name__ == "__main__":
    sampled_frames = sample_youtube_video.invoke({
        "youtube_url": "https://www.youtube.com/watch?v=L1vXCYZAYYM" ,
        "sample_rate": 5
        }
    )

    distinct_species_count = count.invoke({
        "framesStr": sampled_frames
        }
    )

    print(f"Distinct species detected: {distinct_species_count}")