Android-based Fashion Dection in real time

This repository applies transfer-learning-based object detection on Color-Fashion Dataset and deployed the .tflite model on Android to detect in real time.

Download and Requirements

$ git clone https://github.com/leaving-voider/Android-based_Fashion_Dection_in_real_time.git

Python 3.8 or later with all requirements.txt dependencies installed, including tensorflow>=2.4.0 To install run:

$ pip install -r requirements.txt

Example for Custom Dataset

• Download Color-Fashion Dataset

$ cd Android-based_Fashion_Dection_in_real_time

# if needed
# !chmod +x download.sh
$ ./download.sh

$ mv data ../code/data.zip

$ unzip ../code/data.zip > /dev/null 

• Dataset Format Transformation Packing training and testing data into a specific file format which is ".tfrecord", you can do this by run

$ cd ../code

# Transform test data
$ python generate_tfrecord.py --csv_input ../sources/test_labels.csv --output_path test.record

# Transform train data
$ python generate_tfrecord.py --csv_input ../sources/train_labels.csv --output_path train.record

Tutorial Colab

Follow colab for whole process to train your model on custom dataset on lolab