AI Photo / Caption Tool for Site Inspections

[Greenalleycat]

Was thinking about this yesterday on my way to a job...has anyone found a good AI tool for site inspections?
I was thinking of something that goes 'take photo -> voice to text caption -> get back to the office and output a table with photo & caption' but also open to any other tools that people are using

 

[KootK]

Following..

 

[Ron247]

Check into PLAUD. A little practice and diligence in what/how you say something may work. It helped me on projects with a lot of pictures. I use a distinct non-existent word (ChaCha for example) and then my picture description. I had to document every noticeable defect in a property along with its location and what defect I was addressing. Later I globally replace ChaCha with Picture # and then I add the number. Adding the number onsite is too hard for me to keep up with but I may take 150 to 300 pictures at one place, so when you get one number wrong, all the remainders are wrong.

The worst thing about it, is that it records everyone talking at one time, but can to some degree distinguish who they are. I have been in places where 5 people are talking at once and it recorded them all separately.

I am going to monitor this in case anyone has something better, it is needed by many of us.

 

[EngDM]

following

 

[Tomfh]

A builder sent me some excavation photos for new footings. One included a note about a snake in one of the pits. I asked him about the snake, and he said, ‘I haven’t seen any snakes?’ Turns out it was just Apple’s auto-captioning mistaking conduit for a snake!

So make sure to keep an eye on your AI helpers, especially ones that are offering their help quietly in the background.

 

[Ron247]

PLAUD transcribes your recording but it does offer AI summaries and other specialties I rarely use for the same reason. I always proofread what it transcribes and generally have to make corrections but at some point, I either pick different words or it eventually figures some of them out.

I do not use AI to interpret the pictures, I prefer making my own mistakes in that area.

 

[sticksandtriangles]

It's no AI, but I've got some python code that will take a folder of photos and output pages of PDFs with the named photos named based on the .jpg or .png file name like shown below:

Structural_Engineering/ImagestoPDF/ImageToPDF.py at main · retug/Structural_Engineering

Contribute to retug/Structural_Engineering development by creating an account on GitHub.

github.com

You could pretty easily expand this to get a nice cover sheet and add a description below each one to get a professional looking report really quickly.

I've also expanded this to get a full google maps type view with the geodata recorded on each photo that will semi-accurately plot where your photos were taken and allow you to quickly create a photo map.