1

I've trained a model with a custom dataset (Garfield images) with Tensorflow Object Detection API (ssd_mobilenet_v1 model) and referring it in the android sample application available on Tensorflow repository. The application can only detected the images in distances less or equal 20cm approximately.

Do you have any clue about I can improve the model to perform recognitions in longer distances (about 30cm or more) ?

I don't know with this limitation is related with input size I'm using (tested with images with 300x300 and 68x68) or any custom data augmentation is needed to improve that.

Flavio Oliveira
  • 21
  • 4
  • Oliveria Any luck on detecting objects at longer distance? I too could not detect objects at longer distance, please guide me if u have overcome this issue – Dinesh May 09 '19 at 03:05

1 Answers1

0

SSD models are known to have worse performance on small objects. Have you tried using one of our FasterRCNN models to see if the result is acceptable?

Derek Chow
  • 722
  • 3
  • 6
  • In fact, I tried SSD model first because the demo project (android sample) available on tensorflow repository. The sample works fine with ssd mobile net. Do you know if I used a FasterRCNN model on that sample app, it would require many changes on source code to make it work? I'm asking you because that code is kinda of black box for me. Maybe I need to change some input/output nodes from the graph to make it work. Thanks for your help! – Flavio Oliveira Oct 16 '17 at 12:09