I have trained the pet detector from tensorflow object detection with a ssd mobilenet architecture.

The train dataset only contains image with a single objects (1 image = 1 box ).

I would like to detect multiple objects on personnal data. Is it possible to use this model on my data or I need to train a new model with dataset which contains multiple elements ? (for the moment return only 0 or 1 box for 1 image)

More generally, do I need a dataset with images containing multiple objects to have a multiple object detector ?



1 Answer 1


No, you do not need a dataset with many objects on each image in order to train a detector that is able to detect many objects in images. To be more concrete, if you have a dataset with at most one person on each image, it is possible to train a single shot detector that is able to detect several people on an image.

  • $\begingroup$ Thanks for your answer, in addition, if we in the training dataset, there are multiple objects relevant but only one is annoted, is it normal that it gives only one box prediction in my test phase ? $\endgroup$ Commented Dec 8, 2017 at 16:06
  • $\begingroup$ No, this is should not happen in general. However, your detection rate will become worse, since all non-labeled but relevant objects in the training data will basically count as counterexamples. $\endgroup$
    – SaiBot
    Commented Dec 8, 2017 at 16:20
  • $\begingroup$ But if in training phase, models is influenced by only detect one particular object (example : bigger) from a group of object, should not it be normal if the model only return one prediction in test phase ? $\endgroup$ Commented Dec 8, 2017 at 16:40
  • $\begingroup$ Is your original answer is specific to SSD model or it will be true with other detection model (faster rcnn, resnet, inception_resnet, ...) ? $\endgroup$ Commented Dec 8, 2017 at 16:42
  • $\begingroup$ The answer holds for different detection architectures (SSD, Faster rcnn,..) and convolutional networks (inception, mobilenet, resnet, ...) $\endgroup$
    – SaiBot
    Commented Dec 11, 2017 at 8:36

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.