gagolucasm / classify-real-time-desktop

Inception model used to classify camera feed on real time. Coded during the Deep Learning Hackathon 2017 San Francisco

Classify real time desktop and speech

Overview

Team DeepThings (Mez Gebre and I) won the Best Product Category at the Deep Learning Hackathon in San Francisco. We developed in three days a real-time system capable of identifying objects and speaking what it sees, thinking about making a useful tool for the visually impaired, as it could make navigation easier. Proof of concept on a laptop, final model running on Android.

This is only the first prototype for Windows.

The goals / steps of this project are the following:

• Get the Webcam feed without bottlenecks.
• Recognize images using Inception v3.
• Text to speech with Google TTS API.
• Making a functional model.
• Tuninning the parameters.
• Output visual display of the results.

Dependencies

This module requires:

• Python 3.6.1
• Tensorflow-gpu 1.0
• Opencv 3.2
• Numpy 1.12
• Gtts 1.1
• Pygame 1.9

Usage

Just run: python classify_real_time_v2.py

The output should look like this:

More details

For more information, check my medium post here

Licence

This proyect is Copyright © 2016-2017 Lucas Gago. It is free software, and may be redistributed under the terms specified in the MIT Licence.

classify-real-time-desktop's Issues

Help using my only inception PB file

can you help me figure this out?

the keyerror

thank you for the code
and when i run this code,there is an error:
CRITICAL:tensorflow:Failed to locate: n02119789
Traceback (most recent call last):
File "classify_real_time_v2.py", line 167, in
node_lookup = NodeLookup()
File "classify_real_time_v2.py", line 59, in init
self.node_lookup = self.load(label_lookup_path, uid_lookup_path)
File "classify_real_time_v2.py", line 93, in load
name = uid_to_human[val]
KeyError: 'n02119789'