Orbbec Femto Mega Development Guide: Running AI Models Onboard

Last Updated: March 2026

Building with the Femto Mega is surprisingly straightforward—it's basically a Jetson Nano with a camera attached. If you've developed for Jetson before, you'll feel right at home. Here's how to get started.

Setup Options

Option 1: Use the Jetson Directly

The Femto Mega runs a full Jetson Nano OS. Connect a monitor and keyboard, and you have a complete development environment.

Option 2: Development Machine

Develop on your computer, deploy to the Femto Mega. The camera appears as a standard USB device when connected to a host.

SDK Options

Orbbec SDK

Native SDK with full feature access. Works on the Jetson or host PC.

Azure Kinect SDK

If you're coming from Azure Kinect, use the K4A wrapper. Same API—just works.

Python Example

import pyorbbec as ob
import cv2

pipeline = ob.Pipeline()
config = ob.Config()
config.enable_stream(ob.StreamType.DEPTH, 1024, 1024, ob.FPS.FPS_15)
pipeline.start(config)

while True:
    frames = pipeline.wait_for_frames(1000)
    if frames:
        depth = frames.get_depth_frame()
        # Process with AI model here
        # Run TensorFlow/PyTorch directly on Jetson!

pipeline.stop()

AI on the Jetson

Here's what makes the Femto Mega special:

# Run TensorFlow model ON THE CAMERA
import tensorflow as tf

# Load your model
model = tf.saved_model.load('/path/to/model')

# Process depth frame
def process_frame(depth_frame):
    # Preprocess
    input_data = preprocess(depth_frame)
    # Run inference - happens on Jetson!
    results = model(input_data)
    return results

ROS Integration

The Femto Mega works with ROS. Standard topics, familiar workflow.

Conclusion

The Femto Mega gives you full Jetson development capability. If you need onboard AI, this is the depth camera to choose.

Shop: OpenELAB