עבEN
Try Thermal Analysis

Try it

Check your file before you upload it anywhere

The first question in any thermal analysis is whether the file contains measurement data at all. This check runs here in your browser, on the original file, without uploading it anywhere.

Compatibility check

Local file check

The check reads the file structure and reports what is in it: whether radiometric data segments are present, which camera is recorded in the metadata, and the image dimensions. It does not analyse the image and does not compute temperatures.

The file is read in your browser only. It is not sent to a server, not stored, and not shared with anyone.

Drag a file here, or

Original camera file, up to 30MB. Supported: JPG/JPEG.

Analysis on a real image

Element identification on a field thermal image

The images below are real radiometric thermal images. The boxes and values were not drawn by hand: they are the output of running the platform’s identification model over the image at a standard confidence threshold.

Thermal image of a transmission tower with insulator strings, overlaid with component identification boxes.#1267°CDisconnector main contact
Analysis on a real imageThe camera’s own display layer is burned into the image: maximum temperature, ambient temperature, distance and a timestamp. Those are the camera’s values and are not produced by the identification shown here.

Severity by comparative ΔT

The temperatures here were measured from the radiometric data in the file, not read off the burned-in display. Severity comes from the temperature difference against a comparable component of the same type — never from the maximum value alone.

Three starsΔT 243.8°C
Component
Disconnector main contact
Maximum on the hot component
266.6 °C
Comparable reference component
22.8 °C
Comparative ΔT
243.8 °C
Severity
Three stars
Fundamental principle: severity is never set from the maximum value alone. The comparison is made between phases or comparable components on the same structure.

Severity scale

Comparative ΔTSeverity
Below 20°CBelow threshold
20–29°COne star
30–40°CTwo stars
Above 40°CThree stars

Measurement parameters from the file

Emissivity
0.98
Reflected apparent temperature
20 °C
Object distance
50 m

Identified components

The confidence value refers to identifying the component in the image. It is not a fault probability and not a statement about equipment condition.

Run details

Model file
pole_elements.onnx
Architecture
Ultralytics YOLO11n
Input size
512 × 512
Preprocessing
letterbox to 512x512, RGB, /255, CHW
Confidence threshold
0.25
Overlap threshold (NMS)
0.45
Image size
640 × 480
Detections
7
Parity with the application
Not verified
Verified against the cameraInside the measurement box the camera draws on the image, this calculation gives 28.6°C. The camera itself displays Max 29°C on that same image. That agreement is what confirms the decoding is correct.
How the temperatures were derivedThe raw thermal data was extracted from the file and converted to temperatures using the camera’s own calibration constants, corrected for emissivity and reflected temperature. Atmospheric and IR-window corrections were not applied — they affect the absolute value, and far less the difference between two areas of the same image, which is what the rule uses.
This run was performed locally with standard preprocessing, not against the application’s own engine. Until a parity check passes, this result must not be presented as identical to the application’s output.
The files published here were re-encoded with no metadata: no Exif, no location data, and no radiometric payload. The capture date and time were also removed from the picture itself. Measurement parameters are kept, because the analysis cannot be read without them.

Sample analysis

What a result looks like

Sample dataThe sample below is based on a synthetic scene prepared for illustration. It shows the structure of the view — image, findings, measurements, and their source — not an analysis of real equipment.

Suggested findings

Select a finding to highlight the matching region in the image.

Run details

Status
Experimental analysis — sample data
Image source
Synthetic scene prepared for illustration
Reference value
Parallel component marked manually
Engine version
Shown from the actual run once the engine is connected
Limitations
No environmental context; indoor/outdoor classification not shown

The element identification layer

The platform identifies components on the structure, and sub-components within a component — for example the main contact and the blade of a disconnector. These are the component types the platform’s model actually identifies.

  • Structurestructure
  • Insulatorinsulator
  • Clampclamp
  • Disconnectordisconnector
  • Disconnector main contactdisc_main
  • Disconnector bladedisc_knife
  • Cable terminationcable_term
  • Transformertransformer
  • Surge arresterarrester
The identification layer is separate from the thermal finding layer: identification answers “what is this”, a finding answers “what is anomalous here”. Not every identified component leads to a finding.

The boxes here are drawn on a synthetic scene to show the structure. They carry no confidence values, because no model run happened here — a confidence value is shown only from a real run.

Experimental AnalysisThis result supports professional review and does not certify equipment condition.

Service status

Upload for full analysis — next stage

The full analysis environment will open as a limited public trial, once the engine is packaged as a service and passes parity testing against an approved reference version and a golden file set. Until then you will not find an upload button here that returns results: a simulated result is worse than no result.

  • The local check above works today, and reports only what is actually in the file.
  • Full analysis will open first for a verified subset of camera models and files.
  • A file outside the supported set receives an explicit message, not a partial analysis.
Request trial accessTrial Terms

System messages

What the platform will tell you

These are the messages shown when analysis cannot continue. They are written to say what happened and what can be done — not to hide a limitation.

  • No supported radiometric data was found in this file
  • This camera model or file format is not yet supported
  • The file exceeds the allowed size
  • The photo and thermal data cannot be reliably aligned
  • Part of the analysis is unavailable because required data is missing
  • The file and results have been deleted

What happens to your file

  • The local check runs in the browser; the file never leaves your device.
  • When server-side analysis opens, files will be retained for a limited time and deleted automatically.
  • Files are not used for model training by default.
  • Location data and camera serial numbers are not displayed and not sent to analytics.
Full privacy policy

Not sure where this fits in your operation?

Tell us what equipment you inspect and what reporting looks like today, and we will come back with a concrete assessment.

Explore a Pilot for Your OrganizationThermal Analysis