scott huang(黄)

I think what you want is this [GrideSearchCV](http://scikit-learn.org/stable/modules/generated/sklearn.model_selection.GridSearchCV.html#sklearn.model_selection.GridSearchCV), just create a GridSearchCV, pass to pipeline as a "normal " estimator, then you will get what you want.

I think this is a compatibility issues. As I recall older version of gym use > _step , but new one requires > step this is why gym.make will fail...

in order to make sure we are using the same code. I push a v0.1 version a few minutes ago, can you try that version to see if yo still...

you should use multiprocessing version. because performance of multithread version is not good. but if mutithread is working but multiprocess is not, I think maybe this is because the python...

sorry for this. but all my work is done under linux. I don't have time to setup a windows environment to debug this for now. been super busy

you need change your working directory to the project root directory, in my case is : `~/github/object_detection_with_tensorflow` this is python related issues.

One of the tweak of neural network is that you can't guarantee 100% accuracy. if you focus on detecting people, you should retrain network with people only dataset. And collect...

I try to download the DIV2K dataset from official website, **but the competition is ended**, So I don't get a chance to download the pictures, but I check the terms...

@William-Mou Thanks. I can download dataset from this link.

I have the same issue. any fix on that? It is super inconvinient.