Object detection recipe

1.Prepare or download the dataset

Format style, Amount of dataset and labels, Labeling quality, Bbox quality

2.Choose the framework

(1)PyTorch, TensorFlow with keras, mxnet
(1-1)MMDetection, Detectron, SimpleDet

3.Augmentation

DataGenerator, Albumentation, Cut-mix, Mix-up, Mosaic, Insect…

4.Train the models

Faster R-CNN, EfficientDet, Yolo, Recent SOTA model

(Pseudo Labeling)

5.Ensemble

6.TTA

### 1. Introduction
### 2. Data preparation
## 2.1 Load data
## 2.2 Check for null and missing values
## 2.3 Normalization
## 2.4 Reshape
## 2.5 Label encoding
## 2.6 Split training and valdiation set
### 3. CNN
## 3.1 Define the model
## 3.2 Set the optimizer and annealer
## 3.3 Data augmentation
### 4. Evaluate the model
## 4.1 Training and validation curves
## 4.2 Confusion matrix
### 5. Prediction and submition
## 5.1 Predict and Submit results