This project aims to automatically detect failures during 3D printing using computer vision and machine learning. A YOLOv8 model was trained to recognize different types of 3D print failures in real-time. The trained model was then deployed alongside an ESP32-CAM to capture live print images and detect failures, allowing the system to stop or alert the user and prevent material waste.

• Detect 3D print failures like Spaghetti, Layer Shifts,Over Extrusion, Under Extrusion and Stringing.
• Use a YOLOv8 object detection model for accurate detection.
• Deploy the system with ESP32-CAM for real-time inference.
• Improve efficiency and reduce waste in 3D printing.

• Combined multiple publicly available datasets with our oun custom prepared dataset on 3D print failures.

• Annotated images using Roboflow, creating bounding boxes for:

  • Spaghetti
  • Layer_Shift
  • Over_Extrusion
  • Under_Extrusion
  • Stringing
  • good_print

• Exported the dataset in YOLOv8 format.

• Dataset split:

  • Training: 70%
  • Validation: 20%
  • Test: 10%

• Performed data augmentation (flip, brightness, contrast) in Roboflow.

• Model: YOLOv8n (lightweight variant for fast inference)
• Training Command:

  yolo task=detect mode=train model=yolov8n.pt data=data.yaml epochs=50 imgsz=640

• Evaluation Metrics:

  • [email protected]: ~88%
  • Precision: 89.4%
  • Recall: 93.3%
  • F1 Score: 91.3%

[ESP32-CAM] --> captures image -->
[Edge Device] --> runs YOLOv8 inference -->
If failure:
  --> sends signal (Bluetooth/WiFi/UART) to alert/stop print

1. Converted trained YOLOv8 model to ONNX → TFLite.
2. Quantized model using INT8 for lightweight inference.
3. Edge device (e.g., Raspberry Pi or laptop) runs inference.
4. ESP32-CAM captures real-time images and sends them to the edge device.
5. Fault signals (1 or 0) are sent back to ESP32 to take action (e.g., alert/pause).

• Object Detection (YOLOv8)
• Embedded Systems (ESP32-CAM)
• Dataset Annotation (Roboflow)
• Image Processing (OpenCV)
• Edge Deployment (ONNX, TFLite)
• Communication Protocols (UART/Bluetooth/WiFi)
• Evaluation Metrics and Model Optimization

• Store detection history on Firebase / AWS IoT.
• Add a dashboard interface for monitoring.
• Automate printer control with relay modules.

3D-Printing_Failure_Detection-main/
├── data.yaml
├── images/
│   ├── train/
│   ├── valid/
│   └── test/
├── labels/
│   ├── train/
│   ├── valid/
│   └── test/
├── train.py
├── detect.py
├── weights/
│   └── best.pt
├── esp32-code/
│   └── cam_sender.ino
└── README.md