Hand tracker component

[Erol444]

Started integrating @geaxgx's depthai-hand-tracker into the SDK.

TODO:

• [ ] Move HandTrackerRenderer logic into SDK's Visualizer
• [ ] Support 3D hand landmarks (calculation & visualization)
• [ ] Support body pre-focusing
• [ ] Code cleanup
• [ ] Support different model sizes (hand landmarks: lite/full/sparse, perhaps palm_detection_192x192)

Demo

https://github.com/luxonis/depthai/assets/18037362/8d035b2b-2e0f-4f52-8bcf-56a094f62f6f

Code example

from depthai_sdk import OakCamera
from depthai_sdk.components.hand_tracker.renderer import HandTrackerRenderer
from depthai_sdk.classes.packets import HandTrackerPacket
import cv2

with OakCamera() as oak:
    color = oak.create_camera('color')

    handtracker = oak.create_hand_tracker(color)

    render = HandTrackerRenderer(handtracker)
    def cb(packet: HandTrackerPacket):
        render.draw(packet.color_frame, packet.hands)
        cv2.imshow("Hand tracking", render.frame)

    oak.callback(handtracker, cb)
    oak.visualize(handtracker.out.palm_detection)
    oak.visualize(handtracker.out.palm_crop)

    # oak.show_graph()
    oak.start(blocking=True)