Faster RCNN

---

Just a random example, training from scratch, to recognize the labels of the electric diagram. It includes, capacitor,core-air,core-hiron,diode,outlet,resistor,transistor-npn,unspecified,transistor-pnp,earth,battery,core-iron,gate-ampli,transistor-mosfetp,transistor-mosfetn,diodephoto and relay.

Metrics

$$ \begin{align*} precision &= \frac{TP}{TP+FP} \\ recall &= \frac{TP}{TP+FN} \\ F1 &= 2 \frac{recall\times precision}{recall+precision}\\ IOU &= \frac{\text{Intersection area between 2 boxes}}{\text{Union Area between 2 boxes}} \end{align*} $$

Other relevant object detection algorithms

You only Look once, mask rcnn, Haar features classifier ,etc.

References

1. Systems Evaluation Synthetic Documents
2. Elephant Detector Training Using Custom Dataset & YOLOV5
3. Object Detection on Custom Dataset with YOLOv5 using Pytorch and Python
4. yolov5 pytorch ultralytics github
5. Yolo-v5 Object Detection on a custom dataset
6. facebook research detectron2 github

---