Skin Cancer Edge AI Jetson Nano (Demo)

Robotics & Edge Computing Jetson & Embedded Systems Jetson Projects
1

The purpose of this my project (repo: GitHub - gcjordi/skincancer_edgeai_jetsonnano_demo: Skin Cancer Detector - Edge AI classifier - Jetson Nano (DEMO)) is to be a demo for the “Certification Jetson AI Specialist”

It’s a demo (no medical uses at the moment)

It’s a demo for Edge AI Computing based on NVIDIA JETSON NANO 2GB DEVELOPER KIT

Based on the structure: GitHub - dusty-nv/jetson-inference: Hello AI World guide to deploying deep-learning inference networks and deep vision primitives with TensorRT and NVIDIA Jetson.

Dataset: HAM10000 (only part for demo)

Functionality: classify skin lesions (malignant cancer or benign tumor) using Edge AI Computing, with a simple image taken by the user.

P.S: It can also work in WebApp: GitHub - gcjordi/skinlesionanalyzer_webapp_jgc: WebApp para analizador de lesiones de piel mediante ML

RUN (in terminal) (ONLY TEST MODE, NOT REAL MEDICAL TEST):

cd jetson-inference/

docker/run.sh

cd python/training/classification

python3 train.py --model-dir=models/tools --batch-size=4 --workers=1 --epochs=1 data/skin/

python3 onnx_export.py --model-dir=models/skin (previously renamed name folder: tools to skin)

imagenet --model=models/skin/resnet18.onnx --labels=data/skin/labels.txt --input_blob=input_0 --output_blob=output_0 /dev/video0

1 Like
2

Followed your instructions, with some modifications, however I’m unclear of how to run it.

imagenet --model=models/skin/resnet18.onnx --labels=data/skin/labels.txt --input_blob=input_0 --output_blob=output_0 /dev/video0 runs however I don’t see an output when I put in an image.

Also, the python file on github works however it seems setup for the generic model and identifies the skin pictures as ticks. Is there more to the instructions to try this out?

1 Like