via issueshttps://gitlab.com/vgg/via/-/issues2024-01-30T09:13:53Zhttps://gitlab.com/vgg/via/-/issues/448Unable to open .tiff formats2024-01-30T09:13:53ZStephy HuberUnable to open .tiff formatsWould it be possible to add the .tiff format for usage?Would it be possible to add the .tiff format for usage?https://gitlab.com/vgg/via/-/issues/447Some feedback after using VIA3 image annotator2023-12-21T09:35:09ZA GSome feedback after using VIA3 image annotatorI had a lot of fun using VIA3! I want to share some feedback after using the `via_image_annotator.html`.
1. It should not be possible to select any hidden region shape boundaries.
2. When the magnifying glass is enabled, changing visibi...I had a lot of fun using VIA3! I want to share some feedback after using the `via_image_annotator.html`.
1. It should not be possible to select any hidden region shape boundaries.
2. When the magnifying glass is enabled, changing visibility of labels or region shape boundaries is not reflected in the magnifying glass.
![example](/uploads/9cb6344ac55f501d51b3dd85e02d1eb8/example.gif)
3. Region shape boundaries should not display above labels.
4. Allow selection of POINT when it's inside another shape.
5. Labels should not overlap.
![example](/uploads/32d3f2862f83056ea877b9d6035f586f/example.png)
I don't know how to use GitLab but I noticed it says _shiftkey_ instead of shift**K**ey on line 5249 of `via-3.0.12.zip/via-3.0.12/via_image_annotator.html`. I was trying to modify it to only select a shape when holding Shift so that I could start another shape from within an existing shape... but I did not succeed.
6. Ability to reorder and delete options from attribute, updating everything automatically.https://gitlab.com/vgg/via/-/issues/446Cannot add a new attribute after adding 10 attributes2023-12-13T08:39:28ZAbhishek DuttaCannot add a new attribute after adding 10 attributesReported by user Semih
> I've just started using VGG Image Annotator (VIA) 3.0.12. It allows me to add up to 10 attributes. Is it possible to remove or increase this limit?
> After adding the 10th attribute, it sorts the subjects incorr...Reported by user Semih
> I've just started using VGG Image Annotator (VIA) 3.0.12. It allows me to add up to 10 attributes. Is it possible to remove or increase this limit?
> After adding the 10th attribute, it sorts the subjects incorrectly in line 2566.
(see screenshots below)
![Ekran_görüntüsü_2023-12-13_095814](/uploads/91ee8a9ea09d4303d6f3e02c2a229cf2/Ekran_görüntüsü_2023-12-13_095814.png)
![unnamed](/uploads/779f0804caa391bb25168b56e73057e5/unnamed.png)
![unnamed__1_](/uploads/eec082c96b2558118c60aba2496faa14/unnamed__1_.png)https://gitlab.com/vgg/via/-/issues/445Automatic segmentation integration2023-12-11T16:19:31ZMarkAutomatic segmentation integrationOpen-source neural-network assisted segmentation tools are widely available including https://github.com/facebookresearch/segment-anything
It would really speed up workflow, if a reasonable guess at segmentation could be applied to an i...Open-source neural-network assisted segmentation tools are widely available including https://github.com/facebookresearch/segment-anything
It would really speed up workflow, if a reasonable guess at segmentation could be applied to an image before labelling with a toolbox tool to adjust/improve manually.https://gitlab.com/vgg/via/-/issues/444How to load vidoes from a hosted VIA annotator?2023-09-06T05:20:13ZHassan NasserHow to load vidoes from a hosted VIA annotator?I am looking to know how can I tell the via.html I placed on a server to load videos directly from that server. I imagine this should be possible directly as the html and the mp4 are on the same server right?
![image](/uploads/6f3692e55...I am looking to know how can I tell the via.html I placed on a server to load videos directly from that server. I imagine this should be possible directly as the html and the mp4 are on the same server right?
![image](/uploads/6f3692e554fa443336403a05036217af/image.png)
As shown above, I have my html open on a remote link.
Then, I have a file called sample-mp4-file.mp4 on the place where the html is:
![image](/uploads/8195d515488aad1773a491b2abffd457/image.png)
I tried to put: file:///users/rnasser/via-3.0.12/sample-mp4-file.mp4 in the below interface:
![image](/uploads/bf3d1d328e9c4a03509d2f71efa26b76/image.png)
But, things are not working:
![image](/uploads/87b465f3ed47d6a43ab600fbb1d8d239/image.png)
I don't get why. Am I missing something?
Thankshttps://gitlab.com/vgg/via/-/issues/443Annotations erroneously shared between frames in 60 FPS video2023-08-31T10:56:22ZAnne ArchibaldAnnotations erroneously shared between frames in 60 FPS videoI am using the video annotator to add bounding boxes to items in individual frames of a 60 FPS video. Normally I select a region, getting a yellow rectangle, set its attributes, move to the next frame and begin with an empty frame. This ...I am using the video annotator to add bounding boxes to items in individual frames of a 60 FPS video. Normally I select a region, getting a yellow rectangle, set its attributes, move to the next frame and begin with an empty frame. This is the desired behaviour. But sometimes, the yellow rectangle remains in place for the next frame. If I move it around or change its attributes (for example making it appropriate for the new frame) then go back to the previous, the previous frame sees the annotation changed.
I interpret this as the annotation timestamp inadvertently associating it with more than one frame of the video.
This poses a problem as I cannot tell which annotation should be associated with the frame I am working with and which should not; the video contains rapid motion, so annotations valid for one frame are incorrect for the next.
When parsing the annotations, I can write the code that associates timestamps with frames, so I can ensure that the annotations are associated with only a single frame, but it is important that this agree with the interactive annotation process, which appears to have a problem.
I can consider rewriting the videos at 15 FPS (that is, using the same frames but with the in-file FPS set to 15 instead of 60, slowing down the video by a factor of four) but it is not clear to me that this will reliably fix the problem.
I am using Chrome to run the editing, on MacOS.https://gitlab.com/vgg/via/-/issues/440Serving videos from server side2023-06-26T19:20:34ZAbhijat BiswasServing videos from server sideHi -- I have setup the VIA tool as a portion of a larger annotation pipeline for my lab using psiturk. I have a python side script that picks which participants should see which videos. Is there a way to serve a video from the server sid...Hi -- I have setup the VIA tool as a portion of a larger annotation pipeline for my lab using psiturk. I have a python side script that picks which participants should see which videos. Is there a way to serve a video from the server side to the VIA view? From what I can tell the view manager has interfaces for local (client-side) files and remote (public URLs). However, in my case videos are just on the server side. I'm not sure if I'm missing something obvious but ideas appreciated!https://gitlab.com/vgg/via/-/issues/439cursor bug2023-06-30T06:59:12ZEdwin J. Rodriguez Martecursor bugI am facing a problem when using VGG Image Annotator v2.0.12. The problem I am facing happens when I move to the next image to annotate, the cursor changes. Sometimes it changes to a cross with arrows on its four ends, similar to the cur...I am facing a problem when using VGG Image Annotator v2.0.12. The problem I am facing happens when I move to the next image to annotate, the cursor changes. Sometimes it changes to a cross with arrows on its four ends, similar to the cursor used to reposition an annotation. Other times it changes to a cross, but without arrows at its four ends. This happens constantly and makes the annotation process difficult for me.https://gitlab.com/vgg/via/-/issues/437Undo feature in VIA2023-06-12T10:03:49ZAbhishek DuttaUndo feature in VIAHaving an undo feature in VIA will greatly improve the user experience. The Undo feature would be used to undo, for example, an accidentally deleted region or an accidentally added region.Having an undo feature in VIA will greatly improve the user experience. The Undo feature would be used to undo, for example, an accidentally deleted region or an accidentally added region.https://gitlab.com/vgg/via/-/issues/436Region attribute missing after loading annotations2023-06-05T15:55:04ZShufan ZhangRegion attribute missing after loading annotationsI tried to upload annotations in json, coco and csv formats. The region shape attributes loaded successfully but the region attributes didn't. What should I do?I tried to upload annotations in json, coco and csv formats. The region shape attributes loaded successfully but the region attributes didn't. What should I do?https://gitlab.com/vgg/via/-/issues/435Showing keypoint annotations during video playback2023-05-09T05:27:50ZSamuel ZeitvogelShowing keypoint annotations during video playbackI am currently transforming the output of a keypoint detector into a via project file. Using the via_video_annotator, I can now view and edit these annotations. The points are only shown on the video overlay when pausing the video. I tri...I am currently transforming the output of a keypoint detector into a via project file. Using the via_video_annotator, I can now view and edit these annotations. The points are only shown on the video overlay when pausing the video. I tried to add a duration to each keypoint to solve the issue (by adding a second timestamp to the 'z' entry), but this does not seem to work.
Do I have to change the anchor_id (currently "anchor_id":"FILE1_Z2_XY0")?
Thank you for any guidance you can provide.https://gitlab.com/vgg/via/-/issues/434(video) annotation labels vanished when moving to other video file.2023-04-13T09:40:31ZAnshu Kumar(video) annotation labels vanished when moving to other video file.I am using VIA tool for annotating the videos. The steps i'm following are mentioned below.
1. downloading source code via-src-3.0.11.zip (https://www.robots.ox.ac.uk/~vgg/software/via/downloads/via3/via-src-3.0.11.zip)
2. Open **_via_vi...I am using VIA tool for annotating the videos. The steps i'm following are mentioned below.
1. downloading source code via-src-3.0.11.zip (https://www.robots.ox.ac.uk/~vgg/software/via/downloads/via3/via-src-3.0.11.zip)
2. Open **_via_video_annotator.html** file in my default browser (google chrome)
3. added files by click on **local files **
4. After that i added class labels by clicking add button at bottom left most
![Screenshot_2023-04-13_at_3.07.36_PM](/uploads/1451e45e2c0825451b289177dc1e6404/Screenshot_2023-04-13_at_3.07.36_PM.png)
5. But when i'm moving to other video file. These classes vanished
![image](/uploads/05b5fa078429efd88cdf5978aaa759da/image.png)
Could someone help me where i'm doing mistake?
Thank You in advancehttps://gitlab.com/vgg/via/-/issues/433showing annotation dialog after drawing region2023-04-06T11:09:28ZDavid Miguel Susano Pintoshowing annotation dialog after drawing regionWe've had a few emails requesting a feature which we believe is that right after drawing a region, the annotation dialog should be displayed (at the moment, one needs to first draw the region and then click on the region to have the anno...We've had a few emails requesting a feature which we believe is that right after drawing a region, the annotation dialog should be displayed (at the moment, one needs to first draw the region and then click on the region to have the annotation dialog appear).
Here's the exchanged emails for reference:
On Mar 23, 2023 at 3:43 PM Christin Puthur wrote:
> [...] As a part of our project, sonographers in Uganda annotate ultrasound images using the VIA annotation tool. To make the annotation more accurate and easier for the sonographers, some minor tweaks in the VIA tool would be really helpful.
>
> As shown in the screenshot below, we have a set of checkbox options for 'Attribute of a File (e.g. image caption)'. Currently the sonographer checks all the features he can see, and then draws bounding boxes for each of them. The sonographer is then able to select a label for each bounding box from a drop-down menu.
>
> Ideally, we would like the sonographer to be able to check the feature he is viewing and then be prompted to draw a bounding box that would automatically be labelled by the checked feature. This would mean that each checked feature in the image is associated with a bounding box labelled by the checked feature.
>
> Another option would be for the drop-down list of the bounding boxes to only contain features that have been checked in the checkbox options for 'Attribute of a File (e.g. image caption)'. In that case it would be best if the sonographer was forced to label each of the bounding boxes before proceeding to the next image.
![via_tool](/uploads/50751558436148360f5c2c1501a018fb/via_tool.png)
On Mar 23, 2023 at 4:46 PM Abhishek Dutta wrote:
> [...]
> I understand that you have an image that potentially contains multiple organs (e.g. Liver, Spleen, Gall bladder) and you want the human annotators to mark these organs using a bounding box. I am not clear on why two attributes (i.e. "Organ" and "Organ in image") are required because once all the organs in an image have been manually annotated, it is possible to infer all the organs present in the image. I think you work flow is such that you automatically initialise (e.g. using a deep model) each image with a predefined list of organs that is most likely contained in each image (i.e. as a file attribute "Organ") and human annotators have to define bounding boxes corresponding to organs contained in the pre-defined list. Please confirm that I have understood your workflow correctly?
>
> Assuming that I have understood it correctly, we may be able to create a version of VIA that is specifically tuned for this project. The annotator will be presented with a pre-defined list of organs specific to each file and the annotator will be shown a text (e.g. Liver) describing which organ need to be annotated and the box drawn by the annotator will automatically be tagged with this organ name (i.e. Liver). The annotator will be shown next organ in the list and this process will continue until all the organs get annotated in which case it will automatically move to the next image. Will this be sufficient for your requirements?
On Mar 24, 2023 at 9:59 AM Christin Puthur wrote:
> You are right, we do not really need two attributes. But currently we do have them, one associated with the image (a checklist), and another associated with each bounding box (a drop-down menu for selection). The checklist and the drop-down menu have the same list of organs/features.
>
> We do not automatically initialise a predefined list of organs in our workflow. I am sorry if I have not been very clear.
>
> In our workflow, the sonographers simply open an image, see a checklist of possible organs/features (this list is fixed) and mark the ones they are able to see in the image. Following this, they would draw bounding boxes around these organs/features with corresponding labels selected from the drop-down menu. Ideally, we would have the same number of labelled bounding boxes as the number of marked items in the checklist. This workflow does not require the sonographers to type anything for annotation. Also, it helps the sonographers double-check their annotation.
>
> The current problem is that, sometimes, the sonographers create bounding boxes but forget to label them. They might also forget to create a bounding box for a marked feature in the checklist.
>
> To resolve this, one solution we considered was for the sonographers to be prompted to draw a bounding box whenever they mark an organ/feature in the checklist. And this bounding box would be automatically labelled by the marked organ/feature.
>
> Another solution could be for the drop-down menu of the bounding boxes to only contain organs/features that have already been marked in the checklist. In this case, the sonographers first mark all the organs/features they see in the checklist and then go on to draw the corresponding bounding boxes. For this solution, it would be best if the sonographers could not leave any of the bounding boxes unlabelled before proceeding to the next image.
On Mar 27, 2023 at 1:16 AM Christin Puthur wrote:
> It would also be very helpful if the video annotator could be modified similar to our requirements for the image annotator. We would like the sonographers to be able to mark in a checklist the organs/features they see in every frame of the video and then draw bounding boxes around these organs/features. Currently, it is only possible to have one checklist associated with a video. We would like there to be a checklist associated with every frame of the video.
>
> If possible, we would like the display of the 'time segment' in the video annotator to be removed. Also, we would like the video annotator to have the same interface as the image annotator where the sonographers can scroll through the frames. It would also be great if there is a way to hide the 'segment' bars that appear below when annotating a video.https://gitlab.com/vgg/via/-/issues/432replace use of VPS with VFS2023-03-31T13:59:53ZDavid Miguel Susano Pintoreplace use of VPS with VFSVIA connects to a VPS (VIA Project Server) server for the shared project feature. We also a have a VFS (Versioned File Store) which at the moment is only used by LISA.
VPS is specific for VIA projects while VFS is a general implementat...VIA connects to a VPS (VIA Project Server) server for the shared project feature. We also a have a VFS (Versioned File Store) which at the moment is only used by LISA.
VPS is specific for VIA projects while VFS is a general implementation. In practice, they're the same with the main difference being that VFS uses uses the following three entries for version control:
```
"project": {
"shared_fid": "__FILE_ID__",
"shared_rev": "__FILE_REV_ID__",
"shared_rev_timestamp": "__FILE_REV_TIMESTAMP__",
```
while VPS expects:
```
"project": {
"pid": "__VIA_PROJECT_ID__",
"rev": "__VIA_PROJECT_REV_ID__",
"rev_timestamp": "__VIA_PROJECT_REV_TIMESTAMP__",
```
There's no difference between them, it's just the field names and default values.
In addition, VPS development has diverged and implemented checkout specific versions (but I have merged them again so we can have a VFS that does that too).
I propose we change VIA to connect to a VFS server and change the keys above accordingly.
This will cause issues for people with old shared projects running newer VIA versions and maybe we should migrate things from one to the other and have VFS be capable to handle both for a while. Opinions?https://gitlab.com/vgg/via/-/issues/431VIA updates all option-id when a new option is added to an attribute of type ...2023-03-27T08:31:17ZAbhishek DuttaVIA updates all option-id when a new option is added to an attribute of type SELECTAn attribute of type SELECT (i.e. dropdown) has a set of options to define the choices available in the dropdown menu. These choices can be updated by the user using the "Attribute Editor" in the top toolbar. If the user adds a new choic...An attribute of type SELECT (i.e. dropdown) has a set of options to define the choices available in the dropdown menu. These choices can be updated by the user using the "Attribute Editor" in the top toolbar. If the user adds a new choice in the dropdown, the option-id of all other attributes gets changed. This will break all existing annotations. See the illustration below taken from a real manual annotation project which reported this issue.
![attribute-option-update-issue](/uploads/8c6a2edd6671f168228f395e14964cc7/attribute-option-update-issue.png)https://gitlab.com/vgg/via/-/issues/428Can't load image from Google Drive or Dropbox2023-03-09T11:38:07ZZhongqi ZhengCan't load image from Google Drive or DropboxHi, our team is new to VIA, and we tried to load images from public accessible Google Drive/Dropbox folder, but it seemed that VIA can't recognize the URL and load image. We wonder what kind of cloud storage would work. We also appreciat...Hi, our team is new to VIA, and we tried to load images from public accessible Google Drive/Dropbox folder, but it seemed that VIA can't recognize the URL and load image. We wonder what kind of cloud storage would work. We also appreciate any advice on collaborative work. Please let me know if you need any more information. Thanks!https://gitlab.com/vgg/via/-/issues/426coco import fails with empty segmentation2023-01-13T16:57:18ZDavid Miguel Susano Pintococo import fails with empty segmentationCOCO format specifies that a segmentation is a list of polygons, and that list of polygons has a length of multiple of 2 (actually doesn't say that list bit anywhere but it is a list of pair of coordinates). As such, it should be fine fo...COCO format specifies that a segmentation is a list of polygons, and that list of polygons has a length of multiple of 2 (actually doesn't say that list bit anywhere but it is a list of pair of coordinates). As such, it should be fine for any of those lists to be empty.
In my particular case, the detector is finding regions that are effectively empty regions. The output of the detection is in RLE format so there's no coordinates - only an empty mask. When those masks are converted to polygons, they're converted to an empty list. VIA then chokes on them:
```
Uncaught (in promise) TypeError: pts is undefined
polygon_to_bbox https://www.robots.ox.ac.uk/~vgg/software/via/via.html:11145
import_coco_annotations_from_json https://www.robots.ox.ac.uk/~vgg/software/via/via.html:1873
import_coco_annotations_from_json https://www.robots.ox.ac.uk/~vgg/software/via/via.html:1794
load_text_file https://www.robots.ox.ac.uk/~vgg/software/via/via.html:2167
load_text_file https://www.robots.ox.ac.uk/~vgg/software/via/via.html:2166
load_coco_annotations_json_file https://www.robots.ox.ac.uk/~vgg/software/via/via.html:1635
sel_local_data_file https://www.robots.ox.ac.uk/~vgg/software/via/via.html:1574
onclick https://www.robots.ox.ac.uk/~vgg/software/via/via.html:1
```https://gitlab.com/vgg/via/-/issues/425VIA application cannot find my video file2023-01-11T09:21:25ZGUANYUN WANGVIA application cannot find my video fileI am trying to upload my videos on the application but it failed. Every video is about 700-800MB. I have tried some smaller video like 20MB, and it was successful. Does VIA application has limitation on file size?
Should I cut my video i...I am trying to upload my videos on the application but it failed. Every video is about 700-800MB. I have tried some smaller video like 20MB, and it was successful. Does VIA application has limitation on file size?
Should I cut my video into pieces?https://gitlab.com/vgg/via/-/issues/424fails to load COCO annotations with segmentations in RLE2023-01-10T14:23:50ZDavid Miguel Susano Pintofails to load COCO annotations with segmentations in RLEAnnotations in COCO format can be defined as a list of polygons or as RLE mask. From the [format documentation itself](https://cocodataset.org/#format-data):
```
annotation{
"id" : int,
"image_id" : int,
"categor...Annotations in COCO format can be defined as a list of polygons or as RLE mask. From the [format documentation itself](https://cocodataset.org/#format-data):
```
annotation{
"id" : int,
"image_id" : int,
"category_id" : int,
"segmentation" : RLE or [polygon],
"area" : float,
"bbox" : [x,y,width,height],
"iscrowd" : 0 or 1,
}
```
However, VIA 2 (VIA 3 doesn't even load COCO annotations) always assumes that `"segmentation`" is a `[polygon]` (see https://gitlab.com/vgg/via/-/blob/13107d24d9df415a2116c897f0776c9a6d40dbac/via-2.x.y/src/via.js#L870 ).https://gitlab.com/vgg/via/-/issues/423How to import temporal segments saved as a CSV file?2023-01-07T08:33:37ZRafay KhanHow to import temporal segments saved as a CSV file?Hi, there seems to be no option to import temporal segments saved as a CSV file. I would like to load the temporal segments to validate the annotations. Is there any way to load them up?
My CSV file was exported through the `Only Tempor...Hi, there seems to be no option to import temporal segments saved as a CSV file. I would like to load the temporal segments to validate the annotations. Is there any way to load them up?
My CSV file was exported through the `Only Temporal Segments as CSV` option and looks as follows:
```
# Exported using VGG Image Annotator (http://www.robots.ox.ac.uk/~vgg/software/via)
# CSV_HEADER = metadata_id,file_list,temporal_segment_start,temporal_segment_end,metadata
"1_jBqksA2A","[""1_1_crop.mp4""]",5.106,14.53574,"{""TEMPORAL-SEGMENTS"":""interaction""}"
"1_941FpUEG","[""1_1_crop.mp4""]",14.739,25.23944,"{""TEMPORAL-SEGMENTS"":""interaction""}"
"1_AW44Wq7q","[""1_1_crop.mp4""]",30.56,42.73944,"{""TEMPORAL-SEGMENTS"":""interaction""}"
"1_XVZXuGO9","[""1_1_crop.mp4""]",44.527,46.64685,"{""TEMPORAL-SEGMENTS"":""interaction""}"
"1_1tgharPF","[""1_1_crop.mp4""]",52.111,55.9987,"{""TEMPORAL-SEGMENTS"":""interaction""}"
"1_8LZBjdYJ","[""1_1_crop.mp4""]",66.908,75.94315,"{""TEMPORAL-SEGMENTS"":""interaction""}"
"1_8kNfSb9G","[""1_1_crop.mp4""]",78.242,81.10717,"{""TEMPORAL-SEGMENTS"":""interaction""}"
"1_PwUawXYl","[""1_1_crop.mp4""]",86.017,95.41152,"{""TEMPORAL-SEGMENTS"":""interaction""}"
"1_N1wf4WKd","[""1_1_crop.mp4""]",98,100.62891,"{""TEMPORAL-SEGMENTS"":""interaction""}"
"1_YyqgpqnA","[""1_1_crop.mp4""]",109.025,113.67239,"{""TEMPORAL-SEGMENTS"":""interaction""}"
"1_p7loXcZZ","[""1_1_crop.mp4""]",119.895,123.32457,"{""TEMPORAL-SEGMENTS"":""interaction""}"
```
The only import options I see in VIA 3.0.11 (via_video_annotator) are:
![image](/uploads/313d2b6f3c9f6372cce38c1d273cdf34/image.png)
For testing, I'm attaching this CSV file and the corresponding video from the MERL Shopping dataset.
[TestViaProject04Jan2023_19h25m42s_export.csv](/uploads/ffffaff0ecef9d46f4e5bcb0d37ed9c1/TestViaProject04Jan2023_19h25m42s_export.csv)
![1_1_crop](/uploads/723b2857392da8500a5de69dfc0a7e8b/1_1_crop.mp4)