TensorFlowUnity-ObjectDetection icon indicating copy to clipboard operation
TensorFlowUnity-ObjectDetection copied to clipboard

Changed the graph nothing happens

Open christoschatzakis opened this issue 7 years ago • 2 comments

I replaced the graph and the labelmap in the inspector but nothing happens.

I have trained a RCNN model in tensorflow 1.9, exported the frozen_graph.pb, changed it in . bytes etc. I installed the tensorflowsharp (1.7) unity plugin and it gave me an error. I installed the nuget but it made to difference in unity.

So, I exported the graph again in tf 1.7 format and loaded it in unity. No error this time. In the console I get the "Graph loaded" but no detection is happening. (ps. I have tested the graph in python and it works).

Help, please!

christoschatzakis avatar Aug 06 '18 23:08 christoschatzakis

Hi, have you added Display name in label map.txt file?

ragavendranbala avatar Jan 30 '19 02:01 ragavendranbala

Hi, I am suffering from the same problem. I have 2 labels: class 1 and class 2. I have created a protobuf which works in python. I changed it to .bytes, replaced .bytes and meta file in this repo. There is still no detection working. What else do I need to do?

Amanpradhan avatar Feb 08 '19 10:02 Amanpradhan