15th Feb 2024: Astro Pixel Processor 2.0.0-beta29 released - macOS native File Chooser, macOS CMD-Q fixed, read-only Fits on network fixed and other bug fixes
7th December 2023: added payment option Alipay to purchase Astro Pixel Processor from China, Hong Kong, Macau, Taiwan, Korea, Japan and other countries where Alipay is used.
Hi. I'm new to this forum but I've used the APP trial. It took a bit of learning but I did like it. I noticed it brought out more signal in the final image than DSS did. I want to purchase it but I do really want GPU acceleration to work before I jump in. Also saw some mentions of an internal star detection algo. Could that be used to remove stars like Starnet++ or StarXterminator? Send all that to the GPU as well?
@digug Hi George,
Currently APP 2.0.0 is in preparation which focuses on saving settings and intermediate processing state, as well as an extensive user manual.
GPU acceleration is on the TODO list but not for this version. Why do you insist on having that?
There is a Star Reduction tool that allows for reducing the star sizes and in some cases removing the stars completely. It works in a similar way as the other tools that you mention but is a custom implementation and doesn't allow yet to separate the stars, process them separately and recombine again. That, too, is planned for a future version of APP.
HTH, Wouter
GPU acceleration is on the TODO list but not for this version. Why do you insist on having that?
Premium price is why I insist. I enjoyed the trial but the processing time took about as long and in some cases longer than DSS. Once CUDA or GPU acceleration for processing images, not just star removal, becomes a standard supported feature I will definitely make the purchase.
I agree on the GPU acceleration, pretty much every program doing anything with images uses GPU acceleration. Why would you leave all that calculating power unused?
My two cents here is that, typically, Neural network models benefit a lot from GPU acceleration. GPUs perform matrix transforms with very large number of parallel operations and quickly. Matrix transforms depend wholly on multiply-and-accumulate operations that are super optimized in GPUs. AFAIK (and I am not part of the dev team), APP probably does have a lot of signal processing steps and transforms that may not be readily available with standard CUDA libraries (required for NVIDIA GPUs) so development and optimization for speed and memory can take a lot of time. For example, certain spectral transforms or noise estimation algorithms may not be readily implementable as vanilla matrix transforms. Many processing steps algorithmically are quite core and low level in APP, they do not have neural network equivalents for them. Some steps like star removal may use neural models, and I am sure the APP team has a plan for it.
In addition, any stacking software, APP or otherwise, a significant bottleneck is also file I/O speed that GPUs cannot help with. At minimum, a user is processing 100 - 200 Light files that need to be opened, 16/32-bit data loaded and processed. Even a simple operation like pixel normalization, file loading time (that is operating system and disk performance dependent) can be significantly higher than the actual processing on CPUs.
With all this said, even on a modern, but not bleeding edge CPU (my 10th gen i7 is already 3+ years old) APP is very fast and usable. They have good multithreading built in to the processing steps. I also have a 10+year old i7 processor in a Desktop that is quite fast. Most of the speed comes from improving the primary bottleneck, that is File i/o. This is easily tackled with a performant SSDs. That makes a huge difference.
I moved from SiRil &DSS to APP. When I did that transition, I also evaluated PixInsight. PixInsight is a feature rich software, but the learning curve is extremely steep. At that time, I didn't find PixInsight faster in terms of processing time. In fact the manual effort needed to learn the tools far exceeded processing time. APP has worked out of the box and it gets excellent results with very little hands-on tweaking. In my humble opinion, APP is hands down one of the best software out there, for beginners and experts alike.
There are also other benefits of APP: multi-system installation license, multi-os support and in my opinion, accuracy of the result are very good with APP. APP also have excellent intermediate results to inspect and learn from. I feel, as astrophotography is not a cheap hobby, the cost of APP is nothing compared to other tools and equipment and its benefit in hands-off processing makes everything a lot more enjoyable. Hope that is helpful.
Disclaimer: I am just a user. I am not part of the dev team and APP has not requested me to say anything in their support. I came across this post by accident. The above statements are my own opinion after having used APP for 1.5 years.
