Kinect v2 Python Setup Guide (64-bit Fix)

[RichLewi]

Kinect v2 Python Setup Guide (64-bit Fix)

This guide outlines the steps required to get the Kinect for Windows v2 working with Python 3.8+ on a 64-bit Windows system using the pykinect2 library.

1. Prerequisites & Installation

1. Kinect SDK 2.0: Download and install the Kinect for Windows SDK 2.0.
2. USB Port: Ensure the Kinect is plugged into a USB 3.0 port.
3. Python Environment: It is recommended to use a virtual environment (e.g., Conda).

pip install numpy opencv-python comtypes pykinect2 

2. Mandatory Library Patches

Because pykinect2 was designed for 32-bit Python 2.7/3.4, three manual code changes are required within your environment's site-packages folder.

Patch A: 64-bit Memory Alignment Fix

File: .../site-packages/pykinect2/PyKinectV2.py

Location: Line 2216

Action: Comment out the sizeof assertion. On 64-bit systems, this structure is 80 bytes, not 72, which causes a crash.

# Change this:
assert sizeof(tagSTATSTG) == 72, sizeof(tagSTATSTG)

# To this:
#assert sizeof(tagSTATSTG) == 72, sizeof(tagSTATSTG)

Patch B: Comtypes Version Bypass

File: .../site-packages/comtypes/_tlib_version_checker.py

Location: Inside def _check_version

Action: Add an immediate return to prevent the library from crashing due to version mismatches with modern comtypes.

def _check_version(actual, tlib_cached_mtime=None):
    return # Added to bypass version mismatch errors
    # ... rest of function

Patch C: Python 3.8+ Timing Fix

File: .../site-packages/pykinect2/PyKinectRuntime.py

Location: Lines 154, 312, 330, 350, 374

Action: time.clock() was removed in Python 3.8. Replace all occurrences with time.perf_counter().

# Change:
start_clock = time.clock()

# To:
start_clock = time.perf_counter()

---

3. Working Implementation Script

The following script initializes the Kinect, captures Color and Depth streams, and handles the BGRA-to-BGR conversion for OpenCV display.

import cv2
import numpy as np
from pykinect2 import PyKinectV2
from pykinect2 import PyKinectRuntime

# Initialize the Kinect runtime
kinect = PyKinectRuntime.PyKinectRuntime(PyKinectV2.FrameSourceTypes_Color | PyKinectV2.FrameSourceTypes_Depth)

while True:
    # --- Color Frame ---
    if kinect.has_new_color_frame():
        frame = kinect.get_last_color_frame()
        frame = frame.reshape((1080, 1920, 4))
        # Convert BGRA to BGR for OpenCV
        frame = cv2.cvtColor(frame, cv2.COLOR_BGRA2BGR) 
        frame = cv2.resize(frame, (960, 540))
        cv2.imshow('Kinect Color', frame)

    # --- Depth Frame ---
    if kinect.has_new_depth_frame():
        depth_frame = kinect.get_last_depth_frame()
        depth_frame = depth_frame.reshape((424, 512))
        # Normalize and colorize depth for visualization
        depth_8bit = (depth_frame / 4500.0 * 255).clip(0, 255).astype(np.uint8)
        depth_color = cv2.applyColorMap(depth_8bit, cv2.COLORMAP_JET)
        cv2.imshow('Kinect Depth', depth_color)

    if cv2.waitKey(1) & 0xFF == ord('q'):
        break

kinect.close()
cv2.destroyAllWindows()