Dataset '{}' is empty! Error (While train custom dataset on detectron2)

[MuharremcanGulye]

Hello all,
First of all i am almost a newbie.
I am trying to train a model that detects objects on Detectron2.
Problem is all my annotations are csv format and im trying to convert them proper format.
Im trying to imitate baloon detection tutorial.

Here is my code,

import numpy as np
import os, json, cv2, random
import pandas as pd
import cv2

def getdict(path):
files = os.listdir(path)

categories = {'cat1':1, 'cat2: 2, 'cat3':3 }

for file in files:
image_id = int(file.split(".")[0])
dataset_dicts = []
if file[-4:] == '.jpg':
record = {}

  filename = os.path.join(path,file)
  height, width = cv2.imread(filename).shape[:2]

  record["file_name"] = file
  record["image_id"] = image_id
  record["height"] = height
  record["width"] = width

  annotation_file_name = file.split('.')[0]
  annotation_file_path = os.path.normpath(annotation_file_name + '.csv')

  annotation_path = os.path.join(path,annotation_file_path)

  df = pd.read_csv(annotation_path)

  for idx, row in df.iterrows():
    idx += 1
    objs = []
    category = row["raw_classification"]
    obj = {
      "bbox": [row['xmin'], row['ymin'], row['xmax'], row['ymax']],
      "bbox_mode": "BoxMode.XYXY_ABS",
      "category_id" : categories[category]
    }
    objs.append(obj)
    record['annotations'] = objs
    print(record)
    dataset_dicts.append(record)

getdict("/content/drive/MyDrive/test1/")

The error that i get is down below.

/content/detectron2/detectron2/data/build.py in get_detection_dataset_dicts(names, filter_empty, min_keypoints, proposal_files, check_consistency)
262
263 for dataset_name, dicts in zip(names, dataset_dicts):
--> 264 assert len(dicts), "Dataset '{}' is empty!".format(dataset_name)
265
266 if proposal_files is not None:

TypeError: object of type 'NoneType' has no len()

I think i am doing something wrong with creating dataset_dicts but i coudnt find what is wrong.

Thanks and regards.

[rdaggs]

This is an error I am having as well.....Training dataset loads fine, iterates through the epochs.....No such luck with validation dataset.