literate-robot

Application using the model is available at https://literate-robot.azurewebsites.net/ (might take some time to load because of 1.3GB image size) and deployment directory is app

A model which is trained on detecting obstacles from a cleaning robot Point Of View (POV).

Training Notebook: literate_robot.ipynb
Test Notebook: literate-robot-model-test.ipynb
Predictions: literate-robot/PredictedResults
saved_model: frozen_inference_graph.pb
labelmap: labelmap.pbtxt
Labelling tool: LabelImg

Detected classes represents following

fr = furniture(legs/polls)
dr = doors
cb = cables/wires
gr = small garments

Model used: faster_rcnn_inception_v2_coco from Tensorflow 1

Trained using Tensorflow object detection API

Predictions