Hi @BeauNouvelle

Any idea why would I see the following issue: when focusOnFaces is set On I get a strange issue where the image is being seen for a brief moment, then it disappears and the re-apears. It all happens very fast but the issue is noticeable and happens every time.

Any thoughts maybe?

SomeTimes , The project should know the image FaceAware Finished, so I modify the code.

add pass FaceAware Finished image modify.

This is a Notification to pass FaceAwared image value modify, And I think Use block should be delightful. What do you think ?

WeakSelf(imageView) [imageView sd_setImageWithURL:[NSURL URLWithString:url] completed:^(UIImage * _Nullable image, NSError * _Nullable error, SDImageCacheType cacheType, NSURL * _Nullable imageURL) { weakimageView.image = image; weakimageView.focusOnFaces = YES; }];

When I scroll the TableView, This code will be performed.( I used SDWebImage to set ImageView image and Then focusOnFaces )
And the memory report The memory used so high.

What can do to reduce memory ?

Hi,

First of all — cool library, thanks for sharing it!

I’d like to use the Swift 2.3 version, what version should I refer to on my cocoa pods?

Thanks!

As the title says Xcode 9 gives me a nice title purple exclamation mark telling me that FaceAware is using a method where UIImageView.image is not run on the main thread.

This is an attempt to fix #7. I've also updated the example app to enable landscape support so that this fix can be tested.

About the fix:

First, the app loads in portrait mode:

Then, user rotates the device to landscape mode, making the UIImageView's frame change:

• Without the fix:

• With the fix:

If the frame changes aspect ratio, the visible image layer isn't recomputed with the new aspect ratio.

This is easiest to see on with a full-screen view that handles rotation on a long thin device like an iPhone 5, when you rotate from portrait to landscape the preserved pixels from the original image should be different between the two modes.

Fix for #10

When iterating features rect origin is updated like this:

            rect.origin.x = min(oneRect.origin.x, rect.origin.x)
            rect.origin.y = min(oneRect.origin.y, rect.origin.y)

But initial value for rect.origin was zero - this prevented rect.origin.x and y from being >0. This PR fixes it by using first feature's bounds as initial value.

I wanted to test your library out to see if it would work for an app I am developing, but it looks like you built the entire example app in Xcode 8 Beta using iOS 10. You really shouldn't release public libraries under Betas. Would love to see this re-built in the current release version of Xcode.

Cheers!

This feature adds a new property where you can pass in a closure and be notified once the focus process has been completed.

I've also just realized that this process is run TWICE when first setting up. Once for when 'focusOnFaces' Bool is set, and another during layoutSubviews.

Obviously this will significantly affect performance, especially when placed inside tableview cells. As this has been the behavior for a while, I might create another PR addressing this issue rather than doing it here.

1. When focusOnFaces is set.
2. When layoutSubviews is called.

Face detection is already quite expensive on resources, and to double up on that just isn't acceptable, especially with the number of apps out there now using this library.