We Keep Coding

Custom object detection in Java for Android

I have been working on a custom object detection model where I have a set of images and those need to be identified in another image and in real-time using a phone camera. I am wondering whether it is possible to create a custom model without having to use YOLO or other pre-defined models? As well as if I train the custom model in Python using Tensorflow, Keras, is it possible to export that model to use in Java for an android application?
I have heard of usingCustomVision from Azure, but I have to label each image manually, rather than just using the filename of the image, which becomes time consuming when having a large dataset.
Any advice will be highly appreciated.

How to create a video from a series of images with transition

I want to create a video from a series of images and add a transition between the images (e.g., fade in and fade out). I’m looking for a Java solution and not a C++ one (FFmpeg or anything else).
I checked out JCodec and MediaCodec, but none of those support image transitions.

Have you already tried it with OpenCV for Java? OpenCV offers almost everything you need for such a task. To the transitions you would have to worry about then manually with image blending. But that can be solved with OpenCV too. Just take a look at the documentary
EDIT:
It looks like the Java API lacks the appropriate module that can encode a sequence of images. You can fade images in any case. Take a look at the tutorial. This means you would have to crossfade the images with OpenCV and then use another API (e.g. jcodec) to make a video from the images you create.

Visualize Point Cloud in Project Tango from PCL

I would like to do the following with my application in Project Tango:
Read the Point cloud data.
Filter the points using PCL
Segment Points Using PCL RANSAC plane fitting algorithm
Color each segment
Display the point cloud segmented with different
color on the screen
I have reach step 3 now, and my problem is how to display the points. what I need is an output similar to the plane-fitting C++ example provided by Google. I'm using java point cloud example with native code. I need the displaying to ensure of the filter step and the segmenting output.
My problem is that I don't have any idea how should I perform the visualization in Android from PCL?.
Thanks

For simple task like this, pure OpenGL could work. Tango sample code uses a render library called Rajawalli, which is used more for games.

Using Google Maps with Java on a Desktop Application

I want to use Google Maps on my java based application.
My application would overlay images on maps based on latitude and longitude and change position of these images with time. The time interval will be very short so it needs to be efficient.
Currently I can find following solution, but none of these works :
1) Using Qt WebKit through Qt Jambi : This will just open Google Maps in browser over my application, I would not have control of overlaying images on top of the Map, according to the lat/long.
2) Using Google Maps Image API : This would open image, but I would not be able to translate the location in lat/long into coordinates in image, also as the position of images are changing I will need zoom-in and zoom out feature, which will be unavailable here.
Is there an exiting solution which fits my need?
Thanks!

Manipulating an image on App Engine

I'm trying to draw text to an image from Java in the Google App engine, but I've found no methods (since the java.awt.image.* isn't whitelisted) to do it with. I'm also trying to "layer" one image on top of another, which I can't figure out how to do.
Is there any way to draw text on an image or overlay one image over another in App Engine using Java? Has anyone found any other solutions to this?

You have to use Google's Images API instead.
Not sure if it can do all you need it to, but it does include image composition.