| Network Model | dataset | Accuracy (%) | Limitations |
| [7] Faster RCNN, 2018 | Urban garbage from sanitation dpt. | 89 | Garbage or Not; binary classifications |
| [8] CNN, 2019 | Self-prepared dataset | 96 | Trash or not trash; can’t classify the garbage |
| [9] Inception V3, 2019 | Three classes of image from internet | - | Raspberry Pi application but did not clear about their classification task |
| [10] Public Garbage Net, 2020 | Self-prepared data | 96.35 | Only single class or object in an image; captured data in laboratory with static gray background |
| [11] VGG16, 2020 | Web crawled | 75.6 | Web crawled laboratory images with white clear background |
| [12] Mask-RCNN, 2020 | Beijing Municipal Garbage data | 65.8 | Segmentation recognition is worse with complex or similar background color as image color |
| [13] Inception V3, 2020 | Crawled from web | 93.2 | Crawled web data of some garbage-like object but seems not real garbage data. Poor recognition effect on residual waste and hazardous waste |
| [14] Gnet, 2021; Improved MobilNetv3 | Huawei Garbage Classification Challenge Cup dataset | 92.62 | 40 class of garbage data but it seems a typical object detection; images are not like typical garbage as their background says |
| [15] CNN, | - | 90 | Raspberry model for valuables and garbage with robotic arm. Can’t classify to specific garbage |
| [16] Mask-RCNN, YOLOv4, YOLOv4-tiny, 2021 | TACO open dataset | 83, 97.1, 95.2 | Arduino based trash collection but can’t categories or classify the collected trash |