Hi Ting Y., As discussed and agreed I am confirming the details here.
So here is the requirement :
To develop a model to be used in conjunction with an iphone app to detect hands.
- The model needs to be of .mlmodel format (This can be converted from various well known platforms using a python tool provided by Apple)
- The model will help detect hands in the live camera feed and identify the position hands to draw a box around it.
- Model should only detect hands and nothing else. The detection should work with 10 foot.
- The bounding box should consist of hand portion from fingers upto wrist
Specifications (Model Evaluation Parameters)
- Input: Image
- Output: MultiArray (Grid with bounding box data)
iPhone app that needs to detect hands upto wrist. Once hand is detected, it should be marked by a bounding box to show it on the iphone.
As you said you will handle the dataset to train the model which will take 3 days. We do not have any dataset to train the model for hands.
There is a python tool provided by Apple that allows you to convert a model from below options to be converted into .mlmodel
You have suggested your preference to use is Caffe v1.
Here are the milestone details:
1. We will have Demo of .mlmodel file to upload in our app and test in our app once completed we will release- 400 USD
Once final approval and source code is received we will pay another - 400 USD