Yet another cheap thermal imager incoming.. Seek Thermal

[encryptededdy]

I do not believe the Seek images are radiometric JPEGs - instead I think it just saves 2 images, one visual, one thermal. However I imagine with the Seek SDK you should be able to make an app that saves radiometric JPEGs or other metadata.

If they take a pair of images, can you please post up some of the thermal part of the pair? I'd be interested in looking at these.

Personally I don't have a seek thermal, but the thermal images are just the 832x624 images you see everywhere. The app outputs just jpegs, and there is no image blending done between visual / thermal. The only reason the visual image is captured is for future review.

Look at this for example: https://www.eevblog.com/forum/testgear/yet-another-cheap-thermal-imager-incoming/msg673081/#msg673081

The 832x624 images you see there (and in most places online) are straight from the seek. You can use google to just search for images with that resolution: http://puu.sh/j0O1c/e80ebfcf1e.jpg

[Ben321]

I do not believe the Seek images are radiometric JPEGs - instead I think it just saves 2 images, one visual, one thermal. However I imagine with the Seek SDK you should be able to make an app that saves radiometric JPEGs or other metadata.

If they take a pair of images, can you please post up some of the thermal part of the pair? I'd be interested in looking at these.

Personally I don't have a seek thermal, but the thermal images are just the 832x624 images you see everywhere. The app outputs just jpegs, and there is no image blending done between visual / thermal. The only reason the visual image is captured is for future review.

Look at this for example: https://www.eevblog.com/forum/testgear/yet-another-cheap-thermal-imager-incoming/msg673081/#msg673081

The 832x624 images you see there (and in most places online) are straight from the seek. You can use google to just search for images with that resolution: http://puu.sh/j0O1c/e80ebfcf1e.jpg

From my understanding, the sensor size is 206x156. I believe that most thermal focal-plane-arrays output 2-byte-per-pixel grayscale data (their data is usually well over 8-bits-per-pixel, often around 14-bits-per-pixel). Since JPEG format does not support any format other than 1-byte-per-channel RGB data, JPEG is not (I'm sure) the only file format that the Seek app can save as. It must also be able to output TIFF or PNG (both of which support the required 2-bytes-per-pixel grayscale data). Therefore I have been expecting to see in the Google Images search a lot of 206x156 images, particularly of the PNG or TIFF formats. Yet I haven't found any. My assumption had previously been that this was because there must have been some sort of non-disclosure-agreement with Seek, that said you weren't allowed to share unedited pictures with others, as unedited pictures contained the full proprietary data set (all those secret Seek algorithms for temperature calculation) within the files, and they were considering that a trade secret, and therefore only edited copies of the files could legally be shared with anybody (hence the ridiculously large dimensions like 832x624, and the ridiculously wrong format of JPEG, for all the Seek image files found on the net).

[frenky]

You can get raw data from Seek with sgstair desktop application: https://github.com/sgstair/winusbdotnet
And here is info about raw frames that you will get from Seek module:
https://github.com/lod/seek-thermal-documentation/wiki/Frame

[Ben321]

You can get raw data from Seek with sgstair desktop application: https://github.com/sgstair/winusbdotnet
And here is info about raw frames that you will get from Seek module:
https://github.com/lod/seek-thermal-documentation/wiki/Frame

Are you telling me that the Seek app does not save any radiometric data at all whatsoever? All it saves is a JPEG image that is identical to what you see in the preview on-screen image as you are pointing it around at things? Why not? What's the purpose of a thermal imager other than to save RAW data? I wonder if the guys that made this kind of missed the point of thermal imagers? I mean it can't even be explained away by saying that it was the easier option to write JPEGs, because it's not. It's much easier to write a raw byte array to a file than to run a compression algorithm that requires external dependencies like libjpeg.dll. Are you SURE that there's no setting in the app to configure it to use write out raw radiometric data to a TIFF or PNG file?

And by the way, I'm not gonna compile that program. I don't have the right version of Visual Studio to do so (I have VS 2010, but it requires a newer version). And my previous request that somebody else compile it for me and post up the EXE file has so far gone unanswered.

I ended up downloading this http://dumb.ro/files/ThermalView.zip instead. It is a fully compiled EXE file, and runs perfectly (now I just need to buy the Seek thermal camera). I'm not sure if it's better quality than the program you mentioned, but until somebody does the aforementioned favor that I asked for, and compiles a copy of the program that you mentioned, and posts the EXE file here for me to download, I have no way to compare them.

[frenky]

Are you telling me that the Seek app does not save any radiometric data at all whatsoever? All it saves is a JPEG image that is identical to what you see in the preview on-screen image as you are pointing it around at things? Why not? What's the purpose of a thermal imager other than to save RAW data? I wonder if the guys that made this kind of missed the point of thermal imagers? I mean it can't even be explained away by saying that it was the easier option to write JPEGs, because it's not. It's much easier to write a raw byte array to a file than to run a compression algorithm that requires external dependencies like libjpeg.dll. Are you SURE that there's no setting in the app to configure it to use write out raw radiometric data to a TIFF or PNG file?

Yes. There is no option to get radiometric data from android app. And no option for TIFF or PNG.

And by the way, I'm not gonna compile that program. I don't have the right version of Visual Studio to do so (I have VS 2010, but it requires a newer version). And my previous request that somebody else compile it for me and post up the EXE file has so far gone unanswered.

I've build it for you (but this program is only usefull as a starting point for writing your own application for Seek):
https://www.dropbox.com/s/n3d33j94v06ewb3/TestSeek.zip?dl=0

I ended up downloading this http://dumb.ro/files/ThermalView.zip instead. It is a fully compiled EXE file, and runs perfectly (now I just need to buy the Seek thermal camera). I'm not sure if it's better quality than the program you mentioned, but until somebody does the aforementioned favor that I asked for, and compiles a copy of the program that you mentioned, and posts the EXE file here for me to download, I have no way to compare them.

Good choice. That is the best user made application for Seek that I know of.

[Ben321]

How do those programs work? Do they just send the same commands the app does (as a result of having reverse engineered the official app and protocol), and thus allow the PC to truly communicate with the Seek thermal imager? Or do they involve a custom-made 3-connector cable, that allows the camera, smart phone, and PC, to all connect together, and then just have the PC siphon off the image stream as the thermal imager tries to send the image stream to the phone (but let the phone send all the initialization commands and other control commands to the thermal imager so that it operates correctly).

[jadew]

How do those programs work? Do they just send the same commands the app does (as a result of having reverse engineered the official app and protocol), and thus allow the PC to truly communicate with the Seek thermal imager? Or do they involve a custom-made 3-connector cable, that allows the camera, smart phone, and PC, to all connect together, and then just have the PC siphon off the image stream as the thermal imager tries to send the image stream to the phone (but let the phone send all the initialization commands and other control commands to the thermal imager so that it operates correctly).

We sniffed the USB traffic.

Personally I got stuck after sniffing the initialization sequence, because my sniffing setup systematically crashed immediately after, so I ended up using the frame request and the disconnect request as described in sgstair's code.

One thing I noticed is that my initialization routine is slightly different from what sgstair found, which I believe is because I used a newer version of the android app. Most of the initialization sequence is just reading configuration data from the device, which we don't know what it means or how to use it and can probably be safely skipped, but all implementations so far are still reading it anyway and do nothing with it.

There's a lot we don't know about the protocol - maybe Seek will reveal some of that information in the SDK documentation or if the SDK will be robust enough, completely remove the need to directly talk to the device.

Edit:
Attached are 3 images, in both PNG and RAW format.

[Ben321]

How do those programs work? Do they just send the same commands the app does (as a result of having reverse engineered the official app and protocol), and thus allow the PC to truly communicate with the Seek thermal imager? Or do they involve a custom-made 3-connector cable, that allows the camera, smart phone, and PC, to all connect together, and then just have the PC siphon off the image stream as the thermal imager tries to send the image stream to the phone (but let the phone send all the initialization commands and other control commands to the thermal imager so that it operates correctly).

We sniffed the USB traffic.

Personally I got stuck after sniffing the initialization sequence, because my sniffing setup systematically crashed immediately after, so I ended up using the frame request and the disconnect request as described in sgstair's code.

One thing I noticed is that my initialization routine is slightly different from what sgstair found, which I believe is because I used a newer version of the android app. Most of the initialization sequence is just reading configuration data from the device, which we don't know what it means or how to use it and can probably be safely skipped, but all implementations so far are still reading it anyway and do nothing with it.

There's a lot we don't know about the protocol - maybe Seek will reveal some of that information in the SDK documentation or if the SDK will be robust enough, completely remove the need to directly talk to the device.

Edit:
Attached are 3 images, in both PNG and RAW format.

What I'm asking, is will I need my phone to use this PC software? Or will I be able to use a Micro-USB to USB converter cable, and attach the Seek thermal imager directly to the PC, and allow the Seek imager to act as a kind of "thermal webcam" with my PC?

[jadew]

What I'm asking, is will I need my phone to use this PC software? Or will I be able to use a Micro-USB to USB converter cable, and attach the Seek thermal imager directly to the PC, and allow the Seek imager to act as a kind of "thermal webcam" with my PC?

You won't need your phone.

[miguelvp]

What I'm asking, is will I need my phone to use this PC software? Or will I be able to use a Micro-USB to USB converter cable, and attach the Seek thermal imager directly to the PC, and allow the Seek imager to act as a kind of "thermal webcam" with my PC?

You won't need your phone.

Well, he will need the phone to get the latest firmware updates if he doesn't have it and only if it does improve something I guess, but not absolutely necessary and you only need to connect it once per update. Meaning you could borrow someone's phone to get the update.

[miguelvp]

I got an email from Seek with a request for participation in Seekware SDK pre-release.  Maybe innovative 3rd party apps will help Seek against Goliath.

The sign up page has the Windows SDK scheduled for release on 07/15/2015.  Better late than never.  I signed up.

Thanks for the heads up, that made me check my personal e-mail and sure enough there it was

I guess I missed the email on Friday but I did get my link to download the SDK and they are also providing a free OTG PC cable for just supplying your sneaker mail address.

Haven't look at it in detail yet and I'm not sure if the license I accepted prevents me from saying more. I'll check, but not sure I'll have time today to play with it.

[jadew]

Neat, looking forward to your reports.

[miguelvp]

Neat, looking forward to your reports.

Not much joy, I removed the zadig drivers because it didn't work with that, but I can't get win7 to pickup the drivers and they are not included on the sdk.

I did send an email to the sdk support email they sent along, so now I'll have to wait.

I did read the terms in detail and I don't see any NDA so I can say that they are targeting even the Raspberry Pi and eventually even the Arm based one
Of course Linux OSX and both mobile platforms

[frenky]

The windows desktop application from jadew now has many more palettes: 
https://www.eevblog.com/forum/chat/seekthermal-how-to-correct-the-image-received-from-the-sensor/msg713582/#msg713582

[miguelvp]

I did get a reply by this morning and they supplied me the drivers

I got the sample working, the sdk is pretty spartan for now but very useful at the same time.

And by peeking at the library it seems it has a lot of potential, not sure how much of it they will open up.

I don't know how much I'm supposed to keep to myself but this is the output image of my router and if I hover the cursor over where I've placed that circle it shows me that the temperature is 42.904 Celsius and the unsigned integer value is 2990 decimal.



I did use a lens to get more detail btw, and I did crop the image to only show the output (and rotated it since I took it sideways)

This is the output of an older version of my code with the same lens but not at the same distance.

[frenky]

Great, I hope to get access to SDK soon...

[bsliv]

I got an email from Seek with a request for participation in Seekware SDK pre-release.  Maybe innovative 3rd party apps will help Seek against Goliath.

The sign up page has the Windows SDK scheduled for release on 07/15/2015.  Better late than never.  I signed up.

Thanks for the heads up, that made me check my personal e-mail and sure enough there it was

I guess I missed the email on Friday but I did get my link to download the SDK and they are also providing a free OTG PC cable for just supplying your sneaker mail address.

Haven't look at it in detail yet and I'm not sure if the license I accepted prevents me from saying more. I'll check, but not sure I'll have time today to play with it.

I signed up for the SDK just minutes after I got the email but haven't heard anything back from Seek.  Did Seek send you the link?

[miguelvp]

Yes it came from them, the subject for the email was: "Download Seekware SDK for Windows" and came from "Seek Thermal".

Of course I did post a link on my request regarding my app without their SDK and told them that I could probably go further if I had their SDK, also having Medical and GIS image processing experience during the last 25 years probably helped my case as well

Maybe it's in your junk box.

[miguelvp]

Not the output of the app, not even their LUT, got a grey scale image of my hand and noticed that I could see my veins, so I post processed it to get the highest contrast I could, but didn't do a lot of processing, some mild despeckle and contrast adjustment.



Then since I still had the app running I measured the hotter spot and the cooler spot visually and added the values to the image.
32.894 C is the maximum but rounded up to 33, the next highest value was 32.851 C so the step at that range is just 0.043 K on other spots I saw the deltas being 0.044 K increments but didn't analyze it with the data, just moving my cursor around for now.

the maximum integer value was 2819 and the other one was 2817 but I seen those values on other areas of the image so it seems the temperature data is averaged, so the integer values might be unprocessed.

The lower value is 24.478 (2625 int value) but rounded it up to 25 in the picture. Next value near was 24.523 (int value 2626-2628) so it seems the integer has more resolution than what they average for the temperature array for the full image. But still around 0.045 K deltas on the temperature array, but it seems the sensors are 4 times more accurate than that.

Sorry for mixing Celsius for absolutes and Kelvin for deltas.

Also I see vertical banding from their algorithm, they seem to go away if I rotate the camera 90 degrees.

My ring is around 28 C at its coldest spot.

Sorry if I didn't alter the program more to get more accurate values and deltas, but I only have time to play with it for now.

Here is the original pre-processed image. One thing is that if you stop the camera you can still look at the per pixel values with the mouse of the last captured image, pretty handy (no pun intended).

[frenky]

Anything new to report about SDK? 

[miguelvp]

You can select from several LUTs to setup the camera and on a single call you can get the RGB image with the LUT applied, also the unsigned int data (not sure if processed or lightly processed yet) and also the temperature on F,C or K per pixel.

Summer hinders things a bit so I don't have a lot of time because the wife wants to do things.

I'll see if I can squeeze some time to check the images from the unsigned data out of the SDK to see how much pre-processing it's been done to the output.

[frenky]

Tnx.

[Uho]

I made new telephoto lenses. Manual focus adjustment in every. Perhaps they will be working with Flir E4 and Hemavision.

[miguelvp]

Nice Uho!

What lenses does it need and are you planning to sell them?

[Uho]

x3 is composed of two lenses. First ZnSe 25mm. Second Ge12.5mm. The image flips.It works well with Seek.I plan to sell. The prototype for HemaVision too X3. It consists of three lenses. It works with Seek. I think that working with other thermal imagers. He has a latter lens 20mm. We need to test. Who knows the diameter of the lens from Flir E4?  Using a telephoto lens X3 for Seek. Video.
https://youtu.be/kZtH9Nq4AHk