Search results
-
M
How do you detect rodents?
I’m trying to detect toddlers who just learned to open the front door, which are also close to squirrels or rats. -
M
CodeProject.AI Version 2.0
It’s purpose built for AI, and won’t do more general calculations like a GPU would. I’d also recommend the one that fits in your WiFi slot. The USB ones tend to be trouble. -
M
YOLO v8 issue with Coral TPU
Yeah, if it were up to me I’d have a single model that runs everything. Has 80 labels for everything relevant to a IPCam setup. Probably based on the YOLOv8 medium, but that would only run well if you have > 2 TPUs. YOLO’s default labels are honestly mostly irrelevant for security cameras. I’m... -
M
YOLO v8 issue with Coral TPU
I’m interested, but when it comes down to it, the TPU is best at running one model. There is significant overhead swapping the internal cache from one model to another. So ideally we would have one model that does everything for object detection, and then all the remaining models are run on the... -
M
YOLO v8 issue with Coral TPU
Inference time improvements will be very context dependent. Some models perform much better spread across TPUs, others do not. Medium and large models tend to, but only to a point. -
M
YOLO v8 issue with Coral TPU
You probably aren’t going to see much inference time improvements with multi-TPU. It’s still doing the same work with the same hardware. However, you should see 4x the inferences per second. -
M
Coral Accelerator unRAID Docker with CodeProject.AI Guide
Single TPU or multi? -
M
CodeProject.AI Version 2.5
The Coral uses 8-bit integers, so there may be some degradation due to that. It shouldn’t be noticeable, however. I suspect that there has been something else going on. The Coral also only has 8 mb of internal memory. A model such as YOLOv8 medium is 24 mb in size. So all of the model past... -
M
YOLO v8 issue with Coral TPU
I think so? I would expect roughly a 2x difference between the medium and large models, but a much smaller % of the large model will actually fit on the TPU when you're running a single TPU. The TPU will cache ~7 MB of a model. The medium model is 24 MB while the large model is 44 MB. The most... -
M
YOLO v8 issue with Coral TPU
I also see a few YOLOv8 large model runs in there. From the log here, it looks as I would expect. How many CPUs do you have? I notice the parallelism is only '1', which seems low. -
M
YOLO v8 issue with Coral TPU
I'm not completely sure what I'm supposed to see in the log, but it does look like the YOLOv8 model is being used regularly by the end of the log file. It cycles from MobileNet to YOLOv8 to a few MobileNets to YOLOv8. You're welcome to try EfficentDet also. I would expect YOLOv8 to perform the... -
M
YOLO v8 issue with Coral TPU
I would expect much better quality results with YOLOv8 than with MobileNet. It is also running faster than I would expect. So maybe there is something wrong and it is running the wrong model? Do you see anything in the logs about what model is being used? In the next version of the code I want... -
M
YOLO v8 issue with Coral TPU
The multi-TPU code is really just a newer version of the code. So it supports the YOLO models, but should probably be tested more before being the default. What do you mean by sensitive? How is your performance with one TPU and the medium model? -
M
Blue Iris and CodeProject.AI ALPR
I wouldn’t suggest it. The Coral TPU works best when there is a single model loaded onto it and it would make sense to keep the larger object detection models going on it. If you had many TPUs hooked up, it could make sense to have the plate reader running on it. But at the same time it may make... -
M
CodeProject.AI Version 2.5
Here's @tigerwillow1 's night cat scene from page 6. No reliable results, but no false positives in the YOLO models. Sometimes the larger models see the cat, sometimes they don't. More support for the medium YOLOv8 model. The other images posted to page 6 didn't give any reliable results. -
M
CodeProject.AI Version 2.5
Here's another set of test images from @Wattstopper101 's cat image with various models: Same applies as before: Coral TPU, etc. This test image is more interesting because the default COCO label set should identify the cat and car(s). But the cat is viewed from the top down and the cars are... -
M
CodeProject.AI Version 2.5
This pic was posted a few pages ago (page 8) by @tigerwillow1 as being tricky for the model they’re running to get correct. I was curious how all of the different models and sizes would handle a ‘tricky’ ipcam image. -
M
CodeProject.AI Version 2.5
For fun I ran the deer image through a bunch of different models and sizes. Here is a link to the output: Note: All models were run on a Coral TPU with an 8-bit model. There is no label for 'deer' in the COCO label set, so a correct answer is to not see anything in the image. The non-YOLO... -
M
CodeProject.AI Version 2.5
YOLOv8 should be able to run inferences on any computer, I'm running it right now on some testing tasks. The history of what 'YOLO' means/is is in those docs also: https://docs.ultralytics.com/#yolo-a-brief-history Basically, it's just iterations on the 'YOLO' algorithm for detecting objects in... -
M
CodeProject.AI Version 2.5
The IPcam models only come in one size as far as I know. They are based off of the small YOLO model. Another idea: make a collection of ‘false positives’. Images like the one you posted over that give weird results. I can run them through different models on the Coral and we can see the... -
M
CodeProject.AI Version 2.5
Have you tried running the same images through different models? See how YOLOv5 compares to v8 and the custom models? That would be very interesting. Edit: it’s also worth trying the different sizes of models. If YOLOv8-small isn’t giving reasonable results, but v8-medium is working, that would... -
M
New Ryzen 8700G Vs 12700K UV with rtx 4060
You may also want to experiment with adding a few Coral TPUs and testing the performance/power there. They only use a maximum of 2 watts each. The next release of CPAI should have much improved multi-TPU support. If you're interested in this route, don't use the USB TPUs unless you have to. The... -
M
Blue Iris and CodeProject.AI ALPR
You can add my vote for getting it working on old GPUs. Before I read the instructions carefully, I had bought a Quadro P400 -
M
CodeProject.AI Version 2.5
Another idea: an IR flood light to help those night pictures. https://a.co/d/c9apfJm Really depends on how much experimentation you want to do. -
M
CodeProject.AI Version 2.5
Yeah, auto mode at night might be the problem. I'd start with these posts: https://ipcamtalk.com/threads/shutter-speed.55376/post-542049 https://ipcamtalk.com/threads/dahua-exposure-settings-shutter-and-gain.58740/post-592022 https://ipcamtalk.com/threads/optimize-shutter-speeds.59924/post-611608 -
M
CodeProject.AI Version 2.5
Part of the problem here is that all of these animals have spent quite a long time working on How Not To Be Seen, and they're good at it. (It's what keeps them alive, after all.) I can't find the rabbit in your last one and your squirrel is a bit of a mess with the compression artifacts. Are you... -
M
CodeProject.AI Version 2.5
I did a quick search for 'small object detection' and came up with these articles: https://blog.roboflow.com/how-to-use-sahi-to-detect-small-objects/ How large are the ipcam models? Perhaps you can export a custom YOLOv8 model with a 1280x1280 input resolution? The other solution (SAHI) is... -
M
CodeProject.AI Version 2.5
Sounds like progress? From what I'm seeing, the cropped/zoomed images are pretty blurry due to motion. Perhaps you could reduce your shutter speed? What is it currently set at?