pedbrgs/fire-detection

Fire and smoke detection using spatial and temporal patterns.

About me

I received my bachelor's degree¹ in Systems Engineering from Universidade Federal de Minas Gerais (UFMG) in 2019, followed by my master's degree² in Computational Intelligence from UFMG in 2021. Currently, I am pursuing a PhD in Optimization at UFMG while also holding the position of Senior AI Engineer at BTG Pactual. My experiences include problem solving in computer vision, data science, machine learning and optimization. In my free time, I like training Brazilian jiu jitsu, skateboarding, traveling and cooking. Feel free to get in touch!

¹ Bachelor's thesis (in Portuguese)
² Master's thesis (in Portuguese), Video presentation (in Portuguese)

---

Selected publications:

• Adriano Chaves Lisboa, Felipe Luz Barbosa Castro, Pedro Vinícius Almeida Borges de Venâncio: Mine Productivity Upper Bounds and Truck Dispatch Rules. In: Mining, 2023.

• Anderson Chaves Carniel, Pedro Vinícius Almeida Borges de Venâncio, Markus Schneider: fsr: An R package for fuzzy spatial data handling. In: Transactions in GIS, 2023.

• Pedro Vinícius Almeida Borges de Venâncio, Roger Júnio Campos, Tamires Martins Rezende, Adriano Chaves Lisboa, Adriano Vilela Barbosa: A hybrid method for fire detection based on spatial and temporal patterns. In: Neural Computing and Applications, 2023.

• Pedro Vinícius Almeida Borges de Venâncio, Adriano Chaves Lisboa, Adriano Vilela Barbosa: An automatic fire detection system based on deep convolutional neural networks for low-power, resource-constrained devices. In: Neural Computing and Applications, 2022.

For additional publications, please visit my profiles on Google Scholar or ResearchGate.

---

Projects:

• Product development
• Research and development

---

fire-detection's Issues

Is baseline models for Fire only?

I was trying your baseline models (Firenet and Mobilenet). Are they for Fire only and not Smoke?
python baseline.py --video <video_file> --model firenet

How can convert yolov5 to yolov8

Hi, I want to ask how can I convert this code into yolov8 format to use with temporal features after detection? Please tell me how can I do this?

Help with some configuration values

Thank you for your work "A hybrid method for fire detection based on spatial and temporal patterns (2023)".

My question is:

• About test video: Is a video start with "FP" is the positive video (having fire/smoke) and those starting with "VP" is the negative sample?
  The video:
  

• About the result table 3: are you use these following values as provided in the code (or the different ones) to produce this result table?

 + conf_thres=0.25,  # confidence threshold
 + iou_thres=0.45,  # NMS IOU threshold"

Result table:

Thank you for your clarification!

Show fire/smoke detections?

I am running the model with python3 baseline.py --video ex1.mp4 --model mobilenet, but it only shows some bash outputs and no output videos. Does it save the output file somewhere?

1/1 [==============================] - 0s 25ms/step
Time taken =  0.04462027549743652
Prediction: non_fire
1/1 [==============================] - 0s 25ms/step
Time taken =  0.043842315673828125
Prediction: non_fire
1/1 [==============================] - 0s 26ms/step
Time taken =  0.04700660705566406
Prediction: non_fire
1/1 [==============================] - 0s 27ms/step
Time taken =  0.050565481185913086
Prediction: non_fire
Results
     video    network  detected  first_frame  time_avg
0  ex1.mp4  mobilenet      True           10  0.049584

Runtime error on CUDA

I have a hard time running it. I guess because the code is for a few years ago and I get issues with new API & A100 GPU etc.
Is there any plans to test or address issues with newer library versions, new GPUs, etc?

RuntimeError: CUDA error: no kernel image is available for execution on the device
CUDA kernel errors might be asynchronously reported at some other API call,so the stacktrace below might be incorrect.
For debugging consider passing CUDA_LAUNCH_BLOCKING=1.