File size: 2,557 Bytes
d244641 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 | ---
license: mit
library_name: torchvision
tags:
- medical
- mri
- object-detection
- brain-tumor
- faster-rcnn
- pytorch
datasets:
- neuron-m1yxd/brain-tumor-ppo4z
metrics:
- map
pipeline_tag: object-detection
---
# Brain Tumor Object Detection (Faster R-CNN)
An end-to-end object detection model using **Faster R-CNN (ResNet-50-FPN)** trained to detect, localize, and classify brain tumors from MRI scans.
- **Model Architecture:** Faster R-CNN with ResNet-50-FPN backbone
- **Framework:** PyTorch & TorchVision
- **Input:** COCO-formatted brain MRI images
- **Classes (3):** `glioma`, `meningioma`, `pituitary`
---
## Dataset
The model was trained using the **[Brain Tumor Dataset on Roboflow Universe](https://universe.roboflow.com/neuron-m1yxd/brain-tumor-ppo4z)**:
* **Classes:** `glioma` (1), `meningioma` (2), `pituitary` (3)
* **Format:** COCO JSON annotations
* **Augmentations:** Horizontal Flip, Random Brightness/Contrast, Color Jitter via Albumentations
---
## Quickstart: Python Inference
You can run inference using standard `torchvision` and `PIL`:
```python
import torch
import torchvision
from torchvision.transforms import functional as F
from PIL import Image
# 1. Load Model Architecture
NUM_CLASSES = 4 # Background + 3 tumor classes
model = torchvision.models.detection.fasterrcnn_resnet50_fpn(
weights=None,
min_size=800,
max_size=1333
)
in_features = model.roi_heads.box_predictor.cls_score.in_features
model.roi_heads.box_predictor = torchvision.models.detection.faster_rcnn.FastRCNNPredictor(in_features, NUM_CLASSES)
# 2. Download and Load Checkpoint from Hugging Face Hub
from huggingface_hub import hf_hub_download
checkpoint_path = hf_hub_download(repo_id="YOUR_HF_USERNAME/YOUR_MODEL_REPO", filename="fasterrcnn_best.pth")
state_dict = torch.load(checkpoint_path, map_location="cpu", weights_only=True)
model.load_state_dict(state_dict)
device = "cuda" if torch.cuda.is_available() else "cpu"
model.to(device)
model.eval()
# 3. Predict on an Image
CLASS_NAMES = {1: "glioma", 2: "meningioma", 3: "pituitary"}
img = Image.open("sample_mri.jpg").convert("RGB")
img_tensor = F.to_tensor(img).unsqueeze(0).to(device)
THRESHOLD = 0.6
with torch.no_grad():
prediction = model(img_tensor)[0]
for i in range(len(prediction["boxes"])):
score = prediction["scores"][i].item()
if score > THRESHOLD:
box = prediction["boxes"][i].cpu().numpy()
label_id = prediction["labels"][i].item()
print(f"Detected {CLASS_NAMES.get(label_id)} with {score:.2f} confidence at {box}") |