Heic support and make tag #165

helgeerbe · 2021-09-01T10:16:11Z

This should fix #164

jgodfrey

I left a few inline comments.

jgodfrey · 2021-09-01T22:44:19Z

picframe/get_image_meta.py

@@ -91,6 +92,9 @@ def get_location(self):

    def get_orientation(self):
        try:
+            ext = os.path.splitext(self.__filename)[1].lower()
+            if ext in ('.heif','.heic'): # converting heic to Image object will rotate pic implicitly. So orientation is always 1
+                return 1


As I mentioned in the issue discussion, I'm not sure we should record "incorrect" orientation data just because (it'll currently) be fixed when the image is opened. Seems like it'd be better to continue to record "accurate" orientation data here. Then, we could just choose to not act on that data for HEIC images at load/display time.

So, store it correctly here, but essentially "ignore" it later for HEIC data.

Might be better. One advantage is, that you don't have to rebuild the cache to display the heic images in correct orientation.

jgodfrey · 2021-09-01T22:47:06Z

picframe/image_cache.py

@@ -9,7 +9,7 @@ class ImageCache:

    EXTENSIONS = ['.png','.jpg','.jpeg','.heif','.heic']
    EXIF_TO_FIELD = {'EXIF FNumber': 'f_number',
-                     'EXIF Make': 'make',
+                     'Image Make': 'make',


I'm not sure which one is correct here (haven't done any research). I will say that my db currently DOES contain some records where the current make field is defined - so, the EXIF Make property must be defined in some cases...

I checked for Apple, Canon, Sony. I only find "Image Make"

DEBUG:exifread:IFD 0 (Image) at offset 8: DEBUG:exifread: Make: (0x010F) ASCII=SONY @ 146 DEBUG:exifread: Model: (0x0110) ASCII=ILCE-7RM3 @ 152

Never "Make" in EXIF:

DEBUG:exifread:Exif SubIFD at offset 286: DEBUG:exifread: ExposureTime: (0x829A) Ratio=3/10 @ 664 DEBUG:exifread: FNumber: (0x829D) Ratio=8 @ 672

Do you have an example image for me?

See exifread: Note that the dictionary keys are the IFD name followed by the tag name. For example:

'EXIF DateTimeOriginal', 'Image Orientation', 'MakerNote FocusMode'

If you check exiftool tags description you will find the "Group" column:

IFD0 = Image

ExifIFD = EXIF

Do you have an example image for me?

Here's a few examples. Looks like about 70% of my (frame) images have EXIF Make instead of Image Make. However, those images also have Exif Model instead of Image Model. So, with the current (mismatched) code, these have the db make defined but not the model.

Ultimately, we might have to be a bit more clever with some of the info and cascade through a number of tags looking for a definition...

Thanks for the example. Checking the standard I would say, that "Make" belongs to 0th IFD (Image) group. And is not defined for Exif IFD (Exif). So in my understanding your sample images are not correct. You are using Picasa, this Software is no longer maintained, isn't it?

What we can easily do is look first in the correct group and as fallback search in the other. Assuming we are focusing only on "Exif" and "Image" group.

exifread is returning everything anyway:

GetImageMeta__tags:{'Image Software': (0x0131) ASCII=Picasa @ 38, 'Image ExifOffset': (0x8769) Long=46 @ 30, 'Thumbnail Compression': (0x0103) Short=JPEG ...yle) @ 558, 'Thumbnail XResolution': (0x011A) Ratio=72 @ 626, 'Thumbnail YResolution': (0x011B) Ratio=72 @ 634, 'Thumbnail ResolutionUnit': (0x0128) Short=Pixel...Inch @ 594, 'Thumbnail JPEGInterc...angeFormat': (0x0201) Long=642 @ 606, 'Thumbnail JPEGInterc...rmatLength': (0x0202) Long=5620 @ 618, 'EXIF ImageWidth': (0x0100) Short=1911 @ 56, 'EXIF ImageLength': (0x0101) Long=2875 @ 68, 'EXIF BitsPerSample': (0x0102) Signed Short=16 @ 80, 'EXIF Make': (0x010F) ASCII=NIKON...TION @ 328, 'EXIF Model': (0x0110) ASCII=NIKON D70 @ 346, 'EXIF Orientation': (0x0112) Short=Horiz...mal) @ 116, ...}

But I wouldn't do this for other Groups like "Thumbnail"

You are using Picasa, this Software is no longer maintained, isn't it?

I'm definitely not a Picasa user (and, yeah, I think it's been gone for a long time), though I did see it noted in the Image Software tag in one of the samples. I assume Picasa (or some part of it) must still be in use somewhere in the Google image pipeline. I must have downloaded that particular image instance from my Google Photos account instead of locating the original copy locally - that's my guess anyway...

What we can easily do is look first in the correct group and as fallback search in the other. Assuming we are focusing only on "Exif" and "Image" group.

Agreed. That's exactly what I meant with my earlier statement...

Ultimately, we might have to be a bit more clever with some of the info and cascade through a number of tags looking for a definition...

But I wouldn't do this for other Groups like "Thumbnail"

Agreed.

So in my understanding your sample images are not correct.

Yeah, I think I agree with you (although, the EXIF spec doc is not a fun read...)

I'll try to get to the bottom of what's creating my questionable EXIF data, although I suspect it's coming from "ON1 Photo Raw", which is my primary RAW converter and image editor. Although, that's surprising if so, as it's a widely used package with regular updates...

jgodfrey · 2021-09-01T22:49:44Z

picframe/viewer_display.py

                if im is None:
                    return None
+                if pics[0].orientation != 1:


If we wanted to shift the "heic orientation" issue to load/display time, it'd need to happen right here somewhere. Either we wouldn't call __orientate_image() for HEIC images or we'd need to change that method to return the original image unchanged if it were passed a HEIC image (though, it'd need to know the image type, which it doesn't know today).

helgeerbe · 2021-09-02T10:06:18Z

I store the original orientation now in the cache. I will prepare a new pull request

jgodfrey · 2021-09-03T22:29:46Z

picframe/get_image_meta.py

+            else:
+                val = self.__get_if_exist(key)
+
+            if val is None:


This looks like it should work to cascade between the two groups, so maybe this is all we need for now. It might be nice (for future) to have something like get_exif_fuzzy (or similar) that used a dictionary keyed by a generic tag name with the value being a list of prioritized group/tag names to search for a match. So, something like this, for example:

{'make': ['Image Make', 'EXIF Make'], 'model': ['Image Model', 'EXIF Model']}

helgeerbe added 4 commits August 31, 2021 13:51

refactor open image so same code is used in all modules

d1db166

fix heic orientation problem

ea7d078

fix unit test orientation

fdbabf8

fix orentation test for heic

b34039b

helgeerbe requested review from jgodfrey and paddywwoof September 1, 2021 10:16

jgodfrey reviewed Sep 1, 2021

View reviewed changes

Store original orientation in cache.

2ed18b4

helgeerbe closed this Sep 2, 2021

helgeerbe reopened this Sep 2, 2021

helgeerbe linked an issue Sep 2, 2021 that may be closed by this pull request

HEIC Images and Make tag #164

Closed

make Exif Tag search more flexible

617ce85

jgodfrey reviewed Sep 3, 2021

View reviewed changes

helgeerbe merged commit 0687716 into dev Sep 6, 2021

helgeerbe mentioned this pull request Sep 6, 2021

HEIC Images and Make tag #164

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Heic support and make tag #165

Heic support and make tag #165

helgeerbe commented Sep 1, 2021

jgodfrey left a comment

jgodfrey Sep 1, 2021

helgeerbe Sep 2, 2021

jgodfrey Sep 1, 2021

helgeerbe Sep 2, 2021 •

edited

Loading

helgeerbe Sep 2, 2021

jgodfrey Sep 2, 2021

helgeerbe Sep 3, 2021 •

edited

Loading

jgodfrey Sep 3, 2021

jgodfrey Sep 3, 2021 •

edited

Loading

jgodfrey Sep 1, 2021

helgeerbe commented Sep 2, 2021

jgodfrey Sep 3, 2021 •

edited

Loading

Heic support and make tag #165

Heic support and make tag #165

Conversation

helgeerbe commented Sep 1, 2021

jgodfrey left a comment

Choose a reason for hiding this comment

jgodfrey Sep 1, 2021

Choose a reason for hiding this comment

helgeerbe Sep 2, 2021

Choose a reason for hiding this comment

jgodfrey Sep 1, 2021

Choose a reason for hiding this comment

helgeerbe Sep 2, 2021 • edited Loading

Choose a reason for hiding this comment

helgeerbe Sep 2, 2021

Choose a reason for hiding this comment

jgodfrey Sep 2, 2021

Choose a reason for hiding this comment

helgeerbe Sep 3, 2021 • edited Loading

Choose a reason for hiding this comment

jgodfrey Sep 3, 2021

Choose a reason for hiding this comment

jgodfrey Sep 3, 2021 • edited Loading

Choose a reason for hiding this comment

jgodfrey Sep 1, 2021

Choose a reason for hiding this comment

helgeerbe commented Sep 2, 2021

jgodfrey Sep 3, 2021 • edited Loading

Choose a reason for hiding this comment

helgeerbe Sep 2, 2021 •

edited

Loading

helgeerbe Sep 3, 2021 •

edited

Loading

jgodfrey Sep 3, 2021 •

edited

Loading

jgodfrey Sep 3, 2021 •

edited

Loading