Infiray and their P2 Pro - discussion

[ksxx]

The auto update thing is a bit of a pain with the Topdon app. In case of an accidental update the older version is quickly reinstalled though. I tested the versions available on apkpure and the last one that works and is available as an apk (and not apkx) is the Mai 31 version (2.00.003).
Unfortunately the original P2pro app is utterly garbage when it comes to measurements. The point and rectangle measurements and the provided statistics are obviously not correct while the topdon app gives reasonable values. Things get much worse when one tries to analyse existing images. Then the measurement data is completely useless. Also a recent update of the infiray app did not change anything in this respect. If had known this before I would have choosen the topdon or some other camera instead.

[mikepcw]

Code: [Select]

ffmpeg -input_format gray16le -video_size 256x384 -r 25 -i /dev/video0 -vf 'crop=h=(ih/2):y=(ih/2), normalize=smoothing=10, pseudocolor=p=inferno' -pix_fmt yuv420p -preset ultrafast -c:v libx264 -f rtsp rtsp://127.0.0.1:8554/path

How can I add @aurel's postprocessing filters (drawtext and graph) that use ffplay into your rtsp output?
I would like to create a stream for my frigate to consume.

[Bob_McBob]

The auto update thing is a bit of a pain with the Topdon app. In case of an accidental update the older version is quickly reinstalled though. I tested the versions available on apkpure and the last one that works and is available as an apk (and not apkx) is the Mai 31 version (2.00.003).
Unfortunately the original P2pro app is utterly garbage when it comes to measurements. The point and rectangle measurements and the provided statistics are obviously not correct while the topdon app gives reasonable values. Things get much worse when one tries to analyse existing images. Then the measurement data is completely useless. Also a recent update of the infiray app did not change anything in this respect. If had known this before I would have choosen the topdon or some other camera instead.

I'm going to return my P2 Pro because of the app issues. This isn't just a minor bug, it's been a year and they haven't fixed it. I'll have to rethink my options, because clearly you cannot rely on app support for these devices. Annoying because the macro attachment is handy.

[smile]

Attention there... that Sharp Infrared Analyzer 0.6.3 file is ringing every virus scanner bell available in this part of germany.

The Nod32 antivirus is silent, Kaspersky also silent.

[smile]

To me the biggest problem is that infiray does not respond to emails, like I tried 3 different emails. How can one trust such company?

[ksxx]

My p2pro seems to give reasonable temp readings at 20C and above (tested and compared with other thermometers). Today I tried it outside at T~-1C. It gave readings for the houses wall of -30C and the window region of -10C! I checked both the p2pro and the topdon app - same result. Then I tried to measure the freezer (inside the house) with its door open. This should be somewhere around -18C. Also getting -30C here. While the temp readings are wrong, I still get a good image contrast, so the problem seems to be in the conversion function and not the sensor itself. Finally I did a series of measurements warming up a cup of cold water, starting at 1C. See the results in the table below. Clearly the error increases drastically towards 0C. Maybe infiray did confuse 0C vs 0K as the absolute zero.... Anything seems possible. Has anyone else with a P2Pro observed this behaviour or is my camera defective???

| ref | P2 |
| ---- | ---- |
| 1 | -5 |
| 2 | -3 |
| 4 | 1 |
| 5 | 2 |
| 6.5 | 4 |
| 7 | 5 |
| 7.5 | 6.5 |
| 9 | 8.5 |

[jdubs]

I'm getting antsy waiting for iPhone 15 support of the native USB-C P2 Pro. If anyone happens to be at CES this week, would you mind stopping by Infiray's booth and maybe poking them about it a little? Thanks!

[PositiveAltitude]

Well, I think the major problem here is physics  . At low temperatures the IR emission is low and the reflected radiation becomes the dominant factor. I just tried to take a look in my freezer - the numbers are all over the place despite I think everything there is relatively the same temperature. Why? because different materials reflect & emit IR at different rates. Ideally, you should use proper emission values in the sensor setting when you take measures of a specific material. And because the emission depends on temperature exponentially this is crucial at low temperatures. So technically I think the low-temperature limit of the device is just a temperature where errors become too much. Maybe you can have more-or-less good results at -10C but you MUST calibrate for material in this range while in +20C it is not that critical.

[ksxx]

all valid arguments but I think this does not explain the wrong temperature readings. Consider the outside experiment for example. There the whole environment (all heat sources except my body) was around -1C and the window areas a little warmer. No reason here to show -30C. For the open freezer the temperature should be higher if there was much reflection from heat sources around it and not lower.

[G28]

My p2pro seems to give reasonable temp readings at 20C and above (tested and compared with other thermometers). Today I tried it outside at T~-1C. It gave readings for the houses wall of -30C and the window region of -10C! I checked both the p2pro and the topdon app - same result. Then I tried to measure the freezer (inside the house) with its door open. This should be somewhere around -18C. Also getting -30C here. While the temp readings are wrong, I still get a good image contrast, so the problem seems to be in the conversion function and not the sensor itself. Finally I did a series of measurements warming up a cup of cold water, starting at 1C. See the results in the table below. Clearly the error increases drastically towards 0C. Maybe infiray did confuse 0C vs 0K as the absolute zero.... Anything seems possible. Has anyone else with a P2Pro observed this behaviour or is my camera defective???

| ref | P2 |
| ---- | ---- |
| 1 | -5 |
| 2 | -3 |
| 4 | 1 |
| 5 | 2 |
| 6.5 | 4 |
| 7 | 5 |
| 7.5 | 6.5 |
| 9 | 8.5 |

FWIW, I have the Topdon TC001 (which supposedly uses the same internals as the P2 Pro) and wrote my own C/C++ app based from Les Wright's Python script.

https://github.com/92es/Thermal-Camera-Redux

When I tested it on my frozen goods, it measured the same as my freezer thermometer.  Granted, neither are lab grade, but both were in sync. 

The app is free and you are welcome to see if it performs any better than the OEM app.  It has already been run on a P2 Pro.

Also, if you have a known delta curve, I could add an optional calibration filter to the app, similar to microphone calibration filters.

[ksxx]

When I tested it on my frozen goods, it measured the same as my freezer thermometer.  Granted, neither are lab grade, but both were in sync. 

Looks like cool a app - thank you! Since I am on windows and have no real hardware Linux box around I cannot test it at the moment (and I dont know how to compile it on windows). However this triggered me to look at the older python code and I tried to get it to work for me. The simple linear temperature conversion posted here in this thread earlier works well, but again I see the same temperature error at low temperatures as with the manufacturer apps. Looks like both the topdon and the infiray app display the normal (upper) video stream as an image. This image looks ok and has contrast also for cold objects. The raw stream (lower half) clips to -30.2C for everything below ~0C and is completely featureless in the cold areas.
If my camera is not an exception, this may point to an internal problem of the Infiray models but I have to do some more testing here. It would be nice if someone with a P2Pro could test their camera on frozen objects to clarify this issue.

stripped down python code to stream video from the cam and temperature conversion here (using the cv2 package)
https://gist.github.com/ks00x/af520dbba1ada0fbdc5e5d6582b22e55#file-p2prolive-py-L10

[G28]

When I tested it on my frozen goods, it measured the same as my freezer thermometer.  Granted, neither are lab grade, but both were in sync. 

Looks like cool a app - thank you! Since I am on windows and have no real hardware Linux box around I cannot test it at the moment (and I dont know how to compile it on windows). However this triggered me to look at the older python code and I tried to get it to work for me. The simple linear temperature conversion posted here in this thread earlier works well, but again I see the same temperature error at low temperatures as with the manufacturer apps. Looks like both the topdon and the infiray app display the normal (upper) video stream as an image. This image looks ok and has contrast also for cold objects. The raw stream (lower half) clips to -30.2C for everything below ~0C and is completely featureless in the cold areas.
If my camera is not an exception, this may point to an internal problem of the Infiray models but I have to do some more testing here. It would be nice if someone with a P2Pro could test their camera on frozen objects to clarify this issue.

stripped down python code to stream video from the cam and temperature conversion here (using the cv2 package)
https://gist.github.com/ks00x/af520dbba1ada0fbdc5e5d6582b22e55#file-p2prolive-py-L10

I believe the original python script had some bugs in it with respect to temp calculations.  It was not calculating all the temps consistently using both bytes.

You can try this one to see if it behaves any better.

[ksxx]

I believe the original python script had some bugs in it with respect to temp calculations.  It was not calculating all the temps consistently using both bytes.

I never got the original script to run and also not your version 🙁. On my windows installation with cv2 4.8.0 the cap.read() method returns a 1D array if the automatic color conversion is switched off to get the raw data. I needed to reshape the array myself to index the pixels and bytes. However as far as I can see, the  temp conversion is exactly the same as what I did (https://gist.github.com/ks00x/af520dbba1ada0fbdc5e5d6582b22e55)

[G28]

I believe the original python script had some bugs in it with respect to temp calculations.  It was not calculating all the temps consistently using both bytes.

I never got the original script to run and also not your version 🙁. On my windows installation with cv2 4.8.0 the cap.read() method returns a 1D array if the automatic color conversion is switched off to get the raw data. I needed to reshape the array myself to index the pixels and bytes. However as far as I can see, the  temp conversion is exactly the same as what I did (https://gist.github.com/ks00x/af520dbba1ada0fbdc5e5d6582b22e55)

I am running 4.51 of OpenCV on Linux.  Wonder if it is the Windows version or the 4.8.0 version of OpenCV that is causing the different return format.  I guess I will find out sooner or later.

[ksxx]

I finally got the code for the reading of the raw temperature data from the P2Pro jpeg files (native app) working properly. Check out https://github.com/ks00x/p2proviewer
I also made a small web app  to view files and export the data as a csv file. On the github page there is a link to an online version of the app.

[ksxx]

Pizza time! Here is an example of the odd low temperature readings with my P2Pro. The upper half of the image is the normal picture from the app and a line profile cutting the pizza in the center. The values in the graph have no physical meaning. The lower part shows the temperature data extracted from the raw data in the file. The line profile values are in deg C. I did measure with an infrared thermometer as a reference and the plastic wrap at the right side of the pizza was ~ -4C and the unwrapped part was ~ -12C. The temperature in the exposed pizza area is totally clipped to -30C. That is also what the native app shows in this area.

[cynfab]

Well, trying not to be in the realm of the man with 2 watches, Here is what I tried.
Got out my 4 chan thermocouple meter and connected up 4  probes, then taped them to a block of aluminum as a stable thermal mass. I used black electrical tape as it has good emissivity. I then stuck the block & probes into the freezer and let it soak for an hour or so. After a while the 2 identical probes were within .1F of each other. I then put the block under the P2 Pro, wiped the condensation off the tape and took the following pictures.
After 15-20 min the block had warmed up to ~40F, the 2 probes were still within a degree or 2 of the P2 Pro.
This is using G28's software.

[ksxx]

After 15-20 min the block had warmed up to ~40F, the 2 probes were still within a degree or 2 of the P2 Pro.
This is using G28's software.

Thank you! This looks convincing and demonstrates that the P2Pro can do the job also at lower temperatures. I would be happy if the temperatures would be somewhat real but I have to assume that my cam is not working properly. Did you also try the readings with the original app as well?

[cynfab]

No, I did not. I got caught up in the middle of my wife wanting to clean out the freezer. So further tests will have to wait a bit.

[G28]

Well, trying not to be in the realm of the man with 2 watches, Here is what I tried.
Got out my 4 chan thermocouple meter and connected up 4  probes, then taped them to a block of aluminum as a stable thermal mass. I used black electrical tape as it has good emissivity. I then stuck the block & probes into the freezer and let it soak for an hour or so. After a while the 2 identical probes were within .1F of each other. I then put the block under the P2 Pro, wiped the condensation off the tape and took the following pictures.
After 15-20 min the block had warmed up to ~40F, the 2 probes were still within a degree or 2 of the P2 Pro.
This is using G28's software.

Thanks for the test.

So the software + P2 Pro appears to be tracking with your thermal probes ?  If so, that is good news.

FWIW, you can add up to 13 user defined temp locations by clicking on the image in user temp mode in addition to the Min/Avg/Max temps to sample other areas of the project.

[cynfab]

Yes, the P2 Pro tracks my thermocouples well, the setup wasn't perfect but the probes were in as intimate contact with the aluminum block as I could manage, the electrical tape was as good a black body as I could find. I would have liked to get the block colder. It actually warmed up from about 6F to 13F in the 2-3 minutes it took to get from the freezer to my lab setup. I wiped the condensation off the tape without touching it with anything other than the wipe so as to not add any heat to the tape. In the picture you can see that the tape that is not in contact with the block is slightly warmer than the tape the measurement is taken from.
Right now the P2 Pro is measuring 66.4F and the thermocouples 64.6F and 64.4F.
The setup has been running continuously for over 2 hours.

[tankslapper]

I finally got the code for the reading of the raw temperature data from the P2Pro jpeg files (native app) working properly. Check out https://github.com/ks00x/p2proviewer
I also made a small web app  to view files and export the data as a csv file. On the github page there is a link to an online version of the app.

This is really useful. If I could make a suggestion - the ability to set the min/max temperature manually would be quite powerful, as that is something missing from the P2 app. Being able to adjust it afterwards makes would make it less of an annoyance.

[ksxx]

This is really useful. If I could make a suggestion - the ability to set the min/max temperature manually would be quite powerful, as that is something missing from the P2 app. Being able to adjust it afterwards makes would make it less of an annoyance.

Yes I agree - Done! Controls are in the sidebar. However one must be a little careful with the interpretation if the manual range is smaller than full range. The colormap 'blackbody' kind off indicates highlight clipping.

The web app can be found here: https://p2proviewer.streamlit.app/

[tankslapper]

Ah you did it already! That works really well.

It will be quite useful for images which include the clear sky which always pulls the scale way out and makes it hard to see smaller differences, say on the outside of a building. It seems that the native app does quite a bit of image processing to make it look a bit smoother, but the raw data is still quite usable.

[Richard Harris]

Hello, I'm looking for product support on the P2 Pro camera.  I purchased a P2 Pro camera a few months ago, I've been having fun with it but I can't successfully record video. When I hit record video, everything looks fine on my phone, but then when I go to view it later there's no video, only a bunch of solid lines. Any help you can provide would be great, I'm using it with a Pixel 6a phone.