efficientdet_lite

efficientdet lite

Input

• Shape : (1, 320, 320, 3)
• Range : [0.0, 1.0]

Output

• category : [0,79]
• probablity : [0.0,1.0]
• position : x, y, w, h [0,1]

Usage

Automatically downloads the tflite file on the first run. It is necessary to be connected to the Internet while downloading.

For the sample image,

$ python3 efficientdet_lite.py

If you want to specify the input image, put the image path after the --input option.
You can use --savepath option to change the name of the output file to save.

$ python3 efficientdet_lite.py --input IMAGE_PATH --savepath SAVE_IMAGE_PATH

By adding the --video option, you can input the video.
If you pass 0 as an argument to VIDEO_PATH, you can use the webcam input instead of the video file.

$ python3 efficientdet_lite.py --video VIDEO_PATH

Reference

pinto

• PINTO_model_zoo

automl

• EfficientDet

python3 model_inspect.py --runmode=saved_model --model_name=efficientdet-lite0   --ckpt_path=checkpoints/efficientdet-lite0  --saved_model_dir=checkpoints/efficientdet-lite0/tflite --tflite_path=checkpoints/efficientdet-lite0/tflite/efficientdet-lite0.tflite

edgeai

• edgeai-modelzoo

efficientdet_lite1_relu.tflite

Framework

Tensorflow 2.7.0

Netron

pinto (int8 / float)

• efficientdet_lite0_integer_quant.tflite
• efficientdet_lite0_float32.tflite
• efficientdet_lite1_integer_quant.tflite
• efficientdet_lite1_float32.tflite

automl (float)

• efficientdet-lite0_integer_quant_automl.tflite
• efficientdet-lite0_automl.tflite

edgeai (float)

• efficientdet_lite1_relu_ti.tflite