[iOS] added VisionTestApp for testing vision ops

[husthyc]

Stack from ghstack:

• [iOS] added VisionTestApp for testing vision ops #3629 [iOS] added VisionTestApp for testing vision ops

How to build & run the test app via Xcode

• cd to ios/VisionTestApp
• run ./setup.sh (this step requires your wget is pre-installed and make sure your Python torch and Python vision are up to date so that the model is able to be generated)
• open VisionTestApp.xcodeproj using Xcode
• build & run (supports both simulators and real devices)
• check if the benchmark results can be displayed on the view
• if you want to check in some code, make sure you run ./clean.sh before commit so that the model file and static libs will not be committed

[fmassa]

Hi @husthyc

Is the PR ready for review? Could you extend the description in the PR with instructions on how to get the app running?
Also, can it be run on an emulator?

[husthyc]

Hi @husthyc

Is the PR ready for review? Could you extend the description in the PR with instructions on how to get the app running?
Also, can it be run on an emulator?

• Yes, I think its now ready for review. Instructions updated.
• Yes, it can be run on a simulator.

[NicolasHug]

@husthyc I gave this a go and getting the app to run was very simple, thanks!

I noticed that currently a dummy image is used, leading to zero detections. Would it be possible to show the forward pass on a real image in order to have a more robust coverage of the pipeline? Thanks!

[husthyc]

@husthyc I gave this a go and getting the app to run was very simple, thanks!

I noticed that currently a dummy image is used, leading to zero detections. Would it be possible to show the forward pass on a real image in order to have a more robust coverage of the pipeline? Thanks!

Sure, will update accordingly.

[NicolasHug]

Hi @husthyc , just checking in: I see you have requested a review from me but I don't think the PR has been updated since my last comment: #3629 (comment)

Did you get a chance to implement the forward pass on a real image? Thanks!

[NicolasHug]

Thanks for your work so far @husthyc !
I just merged it since the feature freeze is approaching fast and it seems like it's properly working already.

A follow-up PR using real input would be very welcome :)

CC @fmassa

[fmassa]

@NicolasHug note that the PR that you merged (which was sent with ghstack) was not merged in master but in a ghstack branch. You'll need to cherry-pick this commit into a commit that needs to be sent to master

[devbooleans]

@husthyc I am unable to build the VisionTestApp. The libraries fails to build with the below error. Is there any additional build settings to be made to run the source.

Undefined symbols for architecture arm64: "torch::jit::Object::find_method(std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char> > const&) const", referenced from: torch::jit::Object::get_method(std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char> > const&) const in ModelRunner.o "torch::jit::Method::operator()(std::__1::vector<c10::IValue, std::__1::allocator<c10::IValue> >, std::__1::unordered_map<std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char> >, c10::IValue, std::__1::hash<std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char> > >, std::__1::equal_to<std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char> > >, std::__1::allocator<std::__1::pair<std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char> > const, c10::IValue> > > const&)", referenced from: torch::jit::Module::forward(std::__1::vector<c10::IValue, std::__1::allocator<c10::IValue> >) in ModelRunner.o "torch::jit::Module::train(bool)", referenced from: torch::jit::Module::eval() in ModelRunner.o "torch::jit::load(std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char> > const&, c10::optional<c10::Device>)", referenced from: +[ModelRunner run] in ModelRunner.o ld: symbol(s) not found for architecture arm64

Xcode version: 12.5
Deployment Target : 14.1 (iPhone device)

[MaddyThakker]

@husthyc I am unable to build the VisionTestApp. The libraries fails to build with the below error. Is there any additional build settings to be made to run the source.

@devbooleans any updates? I am also facing the same issue. Thanks a lot!

[Sabit-Ahmed]

@husthyc I am unable to build the VisionTestApp. The libraries fails to build with the below error. Is there any additional build settings to be made to run the source.

Undefined symbols for architecture arm64: "torch::jit::Object::find_method(std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char> > const&) const", referenced from: torch::jit::Object::get_method(std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char> > const&) const in ModelRunner.o "torch::jit::Method::operator()(std::__1::vector<c10::IValue, std::__1::allocator<c10::IValue> >, std::__1::unordered_map<std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char> >, c10::IValue, std::__1::hash<std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char> > >, std::__1::equal_to<std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char> > >, std::__1::allocator<std::__1::pair<std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char> > const, c10::IValue> > > const&)", referenced from: torch::jit::Module::forward(std::__1::vector<c10::IValue, std::__1::allocator<c10::IValue> >) in ModelRunner.o "torch::jit::Module::train(bool)", referenced from: torch::jit::Module::eval() in ModelRunner.o "torch::jit::load(std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char> > const&, c10::optional<c10::Device>)", referenced from: +[ModelRunner run] in ModelRunner.o ld: symbol(s) not found for architecture arm64

Xcode version: 12.5 Deployment Target : 14.1 (iPhone device)

facing the same issue

[bruno1308]

For those with problems, go to the Build Phases on XCode and add all the content of the lib folder that was generated on the setup.sh step. It should be next to your project's folder.

It should look like this in the end: