Brain Tumor Detection using YOLOv8

In this project we have used different sizes of the YOLOv8 model to detect and classsify brain tumor in the MRI images

Check out paper publication here

---

Dataset

Link To Dataset
Dataset was obtained from Kaggle. It contains 2176 samples of various clinical circumstances. There 455 samples of "glioma", 551 samples of "meningiomas", 620 samples of "pituitary" and 550 samples of "No Tumor".

Training

Training was done on nano and medium size of YOLOv8 model. The training process was optimised with various choice of Optimizers like Adam, Adamax and RMSprop. Following are the training parameters and results.

Sr No	Size	Epochs	Batch Size	Learning Rate	Optimizer	Momentum	Dropout	Precision	Recall
1	Nano	25	109	0.01	Auto	-	-	90.1	79.1
2	Nano	30	32	0.001	Adamax	0.85	0.5	84.5	80.6
3	Medium	30	32	0.001	RMSprop	0.90	0.2	54.7	48.5
4	Medium	30	32	0.001	Adam	0.90	0.3	84.4	84.7
5	Medium	30	32	0.001	Adamax	0.89	0.4	89.9	86.5

Results

1. Nano model with Auto Optimizer
   
   

2. Nano model with Adamax Optimizer
   
   

3. Medium model with RMSprop Optimizer
   
   

4. Medium model with Adam Optimizer
   
   

5. Medium model with Adamax Optimizer
   
   

---

If you have any query, feedback or suggestion feel free to drop a mail at chetan.mahale0220@gmail.com :)