Towards Smoke and Fire detection via images using Deep Learning on resource constrained embedded devices.
 |
 |
Developer: Ary Naim (anaim@unm.edu)
Advisor: Dr. Marios Pattichis (pattichi@unm.edu)
Deep learning libraries:
- TensorFlow
- Keras
INSTRUCTIONS:
- RECOMMENDED START HERE
For the corresponding Google Collab go to: https://colab.research.google.com/drive/1d6qh1uXNrlptKafsonEZLDvkHJko7wD4?usp=sharing
-
For detailed instructions on using custom data sets for to the PDF doc "Step by Step Guide for YOLOv8 Using Keras With Custom Datasets"
-
If you are uploading custom datesets based on the guide in step you can refer to the code in the zip folder "GeneratorBasedBuilder_my_fire_smoke_dataset".
-
For code only refer to dir "code_only"
References:
- Implementation was heavily based on: https://keras.io/examples/vision/yolov8/
- Adding custom dataset via tensorflow: https://www.tensorflow.org/datasets/add_dataset
- Dataset https://github.com/gaiasd/DFireDataset
- Attemptiong to reproduce this paper however using YOLOv8:
https://link.springer.com/article/10.1007/s00521-023-08260-2 https://www.researchgate.net/publication/361649776_An_automatic_fire_detection_system_based_on_deep_convolutional_neural_networks_for_low-power_resource-constrained_devices#fullTextFileContent