Object detection along Singapore Park Connector Network using Mask R-CNN

[@agentx7138]

Creepy person seen in the bushes 13:00 it's like when alone in the woods we see things that aren't there!

[@thisiskindacool]

This is awesome! Was this done realtime? If yes, could you elaborate how you did it as the paper mentioned that the authors accomplished like 5fps...

[@ZheJin]

Vibhor Singh Nope this was done offline and it took 3 hours on an NVIDIA Titan X

[@thisiskindacool]

Oh, i see! Thanks!

[@piyush9555]

@@ZheJin thought it was realtime, how amazing it woule be when realtime

[@HeriRakotomalala]

Did you try on a very busy road where there are more objects and also more occlusion ?

[@Ryo-hi9hi]

How many and What kind of GPU do you use during run?

[@dlam626]

Hi Jin Zhe, did you label your data for training yourself? I am planning to apply it to satellite images

[@ZheJin]

Hi the training dataset is COCO. cocodataset.org

[@taharabdurahmenainouz1037]

Hi. If possible how to train with own datasets (image satellite )

[@杨凡凡-o1m]

I want to classify body movements. What are your ideas?

[@ZheJin]

Hi fanfan, I'm guessing you have a bunch of videos that you would like to infer human actions from. If that's the case: • I've created a list of some recent datasets for activity/action detection here: gist.github.com/jin-zhe/3a6054e99162bc9277940867f942bba2 • You must have a specific set of actions in mind. I would recommend first identifying the dataset that matches your criteria then looking for existing work done by other researchers on that dataset • Once you have identified a research work, you may try to get the pre-trained model from the authors and loading them to run inference on your video data • Of course you will need some prior knowledge in computer vision and AI along with familiarity with tools such as Caffe, C3D, Tensorflow, PyTorch etc. Else if you are a researcher who is keen on this area: • Then you are touching upon an active area of research on "action detection/recognition", "spatial-temporal action localization", "video/scene understanding" • Google those terms and start reading papers ;)

[@prince93600]

You can see recents works done in the field here github.com/jinwchoi/awesome-action-recognition