Dinosaur Classification

A machine learning model for native mobile devices to perform dinosaur image classification using TensorFlow Lite. This ML model is integrated with flutter environment for mobile application. Flutter DinoFinder The code is based on Google Tensorflow model (The link is in Reference section).

Image scraper:

To collect training data-set for dinosaur image classification, the image scraper that efficiently collects image data is implemented. The image scraper link is as follows: Google Image Scraper

Train the network:

For the dinosaur classification problem, it uses mobilenet provided by Tensorflow. In the terminal,

IMAGE_SIZE=224
ARCHITECTURE="mobilenet_0.50_${IMAGE_SIZE}"

python -m scripts.retrain \
  --bottleneck_dir=tf_files/bottlenecks \
  --how_many_training_steps=500 \
  --model_dir=tf_files/models/ \
  --summaries_dir=tf_files/training_summaries/"${ARCHITECTURE}" \
  --output_graph=tf_files/retrained_graph.pb \
  --output_labels=tf_files/retrained_labels.txt \
  --architecture="${ARCHITECTURE}" \
  --image_dir=data

Test the network:

In the terminal,

python -m scripts.label_image \
    --graph=tf_files/retrained_graph.pb  \
    --image=data/Dimetrodon/Dimetrodon_0001.jpg

Note - Create tflite model:

Change the file name from optimized_graph.lite to optimized_graph.tflite

Reference:

TensorFlow For Poets