Missing category label about grocerydataset

[gulvarol]

I am sorry I don't understand your question. The annotation file (https://github.com/gulvarol/grocerydataset#annotationtxt) has the 'brand' id which corresponds to the category. You can convert the information in any format you like.

[CelineZhou03]

Thank you for your reply.
Each <b_i> in this annotation.txt has a value of zero. I want to turn it into a trademark name for each product. in other words，I want to convert annotation.txt annotation information into the format of the .xml file in the annotation inside the VOC2007 dataset. Because I want to use faster r-cnn to train your data set.

[gulvarol]

Zero is the 'other' class. Not all of the class annotations are zero.

[jha-prateek]

Hi, there's a lot of mis-classification in annotation file.

For example look at line 164 in Annotations.txt
File name: C2_P04_N3_S4_1
Cat Ids: ['0', '0', '0', '0', '0', '0', '0', '0', '0', '0', '0', '0', '0', '0', '0', '0', '0', '0', '0', '0', '0', '0', '0', '0', '0', '0', '0', '0', '0', '0', '0', '0', '0', '0', '0', '0', '0', '0', '0', '0', '0', '1', '1', '1', '1', '4', '4', '4', '4', '4', '4', '4', '10', '10', '10']

First shelf of planogram contains all Marlboro(cat id: 1) but in annotations.txt contains all 0.

[gulvarol]

There could be bounding boxes that are not annotated as one of the brand categories:

So among the annotated ones, there should be no wrong labels (they are all manually verified). Among the remaining ones (0: "other" class), there could be missing labels because not every bounding box is exhaustively annotated.

[jha-prateek]

So, how to figure out which of the images are properly annotated?
Because, I see lot of them are not.

[gulvarol]

As I said, the bounding boxes that are annotated as non-zero are manually verified. The 'other' category has some noise, but is largely correct. You can easily annotate more yourself either manually or semi-automatically, as the number of categories are very small.

[jha-prateek]

Ok, Thanks.

[sayakpaul]

How is the csv file structured? Am I right to assume it has the following structure?

@gulvarol

[sayakpaul]

I guess this should be the case -

Could you confirm? @gulvarol

[gulvarol]

I have not created the csv file, it was commited by @revantt

[revantt]

I guess this should be the case -

Could you confirm? @gulvarol

Hi @sayakpaul the headers for this file is [Image_id,x_min,y_min,x_max,y_max,label]

[sayakpaul]

Yeah got it. Maybe updating the csv to include the headers would help reduce this.

[sainivedh19pt]

The labelling of classes in annotations.csv in totally wrong. Surprised by how none has raised it yet