Monocular Depth Estimation using Diffusion Models

Here a little bit more about AI depth map generation. Google published an interesting paper: Monocular Depth Estimation using Diffusion Models   ZoeDepth: Combining relative and metric depth (Official implementation)  

7 methods to get a depth map without exiftool

Online Tools Use an online extractor e.g.: Depth Map Extractor or Depth Extractor (both don't work with my example file from the Google Pixel 6a) Use Photopea. Photopea is even able to import all other images inside the metadata. You can use this tutorial or directly go to .   Command line Tools Depth-map-extractor an Python script optimised for photos from Huawei  phones. Phonedepth , a small command line tool to extract depth data. Creating not extracting In 3D tools like blender you can simply render the mist pass/depth map and then use it in other programs: AI: There are several AI depth map creation tools. (online and local). The quality varies quite a lot depending on the input image and the used model. Sometimes it looks good at first glance, but is simply wrong. However, the AI models are improving very quickly. But if it is not perfect you can fix it, have a look at the last point. Here for example the not very good output result from 3dphot

How to extract mp4 videos from your Google Pixel 6a photos

The Google Pixel 6a phone can take so-called motion pictures. These are jpeg images with a short mp4 video included in the metadata. This is how you see the image in an normal image viewer. This command allows you to extract the video from the jpeg file: exiftool -b -p "${trailer;s/.*(\0\0\0\x1cftypisom)/$1/s}" -ext jpg -w mp4 DIRECTORY or FILENAME -b Output metadata in binary format.  -p "${trailer;s/.*(\0\0\0\x1cftypisom)/$1/s}" Some magic to remove an offset in the binary file. -ext jpg Process files with the specified extension -w mp4 New filename extension DIR Directory or filename    This is the extracted video:  

Depth Map from Google Pixel 6a Photos

An other phone an other method to get the depth map. 😕 If you shoot with an Google Pixel in the "Portrait" mode it saves three images in one jpeg. The First one ist the Main image you See in every image viewer, second one is the image without any filter applied and the third one is the depth map. Main Image You can get the images with the command line tool " exiftool ". This time the depth map is not saved as embedded data but as a trailer. To extract it you need to fist run: "exiftool(-k).exe" MainImage.PORTRAIT.jpg -trailer -b > ImageWithoutFilter.jpg Image without filter This extracts an image with without any filter applied. Now you need to run the same command with the new created image, this will extract now the depth map: "exiftool(-k).exe" ImageWithoutFilter.jpg  -trailer -b > DepthMap.jpg Depth Map    

Things Libre Arts didn't blog about 2022

Alexandre Prokoudine does a great job with his Libre Arts blog and he made a big overview of the year 2022 with a preview of 2023 : As he looks at the big picture I collected some news from small more hidden open source graphics projects: Mobile As I'm only interested in the non mainstream mobile stacks I'm ignoring Android an OS here completely. Megapixels Megapixels got two new releases (1.5 and 1.6) with nice improvements.Meanwhile is Purism working on an fork called millipixels that is using libcamera. libcamera The first version of libcamera released and a lot of big companies like RedHat, Purism and the Raspberry Pi Foundation are working together to improve the camera stack. Image Viewer Image viewers are the kind of programmes that every one thinks are easy to write. But then you realise not only there are thousands of different file formats with metadata (with informations like "rotate this image", "this is an panorama", "geo locations"

Timelapse Czech Republic 2017

  DE:  Ich habe "alte" Bilder aufgeräumt und bin dabei auf eine Zeitraffer-Serie aus dem Jahr 2017 gestoßen. Wenn mich die Metadaten nicht täuschen, habe ich etwa alle 1-2 Sekunden ein Foto gemacht.  Leider sind die Fotos von meinem Handy und nur als JPEG, so dass es in darktable nicht viel zu verbessern gab.  Im Gegensatz zu meinem ersten Versuch mit Zeitrafferaufnahme 2014 habe ich dieses Mal deutlich mehr Kdenlive verwendet, da sich der Funktionsumfang deutlich verbessert hat. (Title, zoom effect, slow playback of the balloon).  Verbesserungsmöglichkeiten:   Mehr Fotos pro Minute Autofokus ausschalten "Richtige" Kamera benutzen und RAW Fotos machen Musik von JuliusH von Pixabay . EN:  I was cleaning up "old" images and came across a time-lapse series from 2017. If the metadata doesn't fool me I took a photo about every 1-2 seconds. Unfortunately the photos are from my cell phone and only as JPEG, so there wasn't much to improve in darktab

Depth Map from Samsung Gallaxy M33 Photos

If you shoot with an Samsung Galaxy M33 in the "Portrait" mode Samsung saves three images inside one jpeg. The first one is the one you see if you open the photo with any application. The second one is the image without any filter applied and the third one is the shot from the depth map camera. Main Image You can get the images from all cameras with the command line tool " exiftool ". First you can check with this commands, if the image contains any additional images:   "exiftool(-k).exe" 20220921_091253.jpg "-embedded*" -a Embedded Image Name : DualShot_1 Embedded Image      : (Binary data 2449408 bytes, use -b option to extract) Embedded Image Name : DualShot_2 Embedded Image 2    : (Binary data 576210 bytes, use -b option to extract) -- press ENTER -- And with this command you can check for a depth map : "exiftool(-k).exe" 20220921_091253.jpg -"*depth*" Depth Map Name      : DualShot_DepthMap_1 Depth Map Data      : (