search

AbnerHqC / GaitSet

A flexible, effective and fast cross-view gait recognition network
577 stars 170 forks source link

creating our own test data(silhouette) #31

Open prb2307 opened 5 years ago

prb2307 commented 5 years ago

@AbnerHqC Thanks for sharing your excellent work. If I want to test GaitSet on my own data, how should I create the silhouette? I was wondering how to input my own data to the network and test the accuracies. If I want to input the video, what should be the approach? (is it the same as creating the silhouettes from the video and using it as an input) Sorry if my questions are too basic. Thank you.

AbnerHqC commented 5 years ago

Hi! Yes, it basically contains two steps to create a new test dataset. First, use a segmentation model to generate silhouettes sequence for each person. Second, use pretreatment.py to crop and align the silhouettes.

prb2307 commented 5 years ago

@AbnerHqC Thanks for the reply. I have gone through the paper. I was a little bit confused between the probe samples and gallery. If I add the test data and change the dataset_path from the config.py to my test data, how does it processes? Does it mean that the test data I have created gets added into the gallery and the probe?
Sorry if my understanding has gone terribly wrong.
this what I found about the gallery and probe : Gallery is one of the data partitions in an algorithm-level biometric evaluation experiment. It is a collection of biometric templates that form the search dataset. Typically, these are representative of the enrolled templates in an actual biometric deployment scenario. In algorithm-level evaluations, care should be taken to have same number of representative templates per subject in the gallery. Probe is the second data partition in an algorithm-level evaluation experiment. It is a collection of biometric templates that need to be recognized or identified by matching against the gallery. In any given algorithm-level evaluation, the probes and gallery differ with respect to the covariate that is being studied. For example, to study the impact of viewpoint covariate, the gallery is chosen to be from one viewpoint and the probe is chosen to be from a different viewpoint.

AbnerHqC commented 5 years ago

@prashant-bansod Hi~ Sorry for the late reply. Yea, the definition of gallery and probe is right. In short, prob is the data as an example and gallery set consists all the candidate samples. The task is to find right sample from gallery set that has same id with the prob data. So to make the test script fit your own dataset, you need to first define the prob and gallery in your dataset and custom codes in utils/evaluator.py.