Detect objects as bounding boxes or masks using one of 28 models from Tensorflow Models repository!
https://github.com/tensorflow/models/blob/master/research/object_detection/g3doc/installation.md
You can download manually from the models zoo: https://github.com/tensorflow/models/blob/master/research/object_detection/g3doc/detection_model_zoo.md
Or you can use the download script provided in this repo!!!
It will download all models for you:
python scripts/download_all_models.py
or you can use ROS to run it:
rosrun object_detection_tensorflow download_all_models.py
This script will download 28 models for you (Updated on 2018.09.16):
- faster_rcnn_inception_resnet_v2_atrous_coco_2018_01_28
- faster_rcnn_inception_resnet_v2_atrous_lowproposals_coco_2018_01_28
- faster_rcnn_inception_resnet_v2_atrous_lowproposals_oid_2018_01_28
- faster_rcnn_inception_resnet_v2_atrous_oid_2018_01_28
- faster_rcnn_inception_v2_coco_2018_01_28
- faster_rcnn_nas_coco_2018_01_28
- faster_rcnn_nas_lowproposals_coco_2018_01_28
- faster_rcnn_resnet101_ava_v2.1_2018_04_30
- faster_rcnn_resnet101_coco_2018_01_28
- faster_rcnn_resnet101_kitti_2018_01_28
- faster_rcnn_resnet101_lowproposals_coco_2018_01_28
- faster_rcnn_resnet50_coco_2018_01_28
- faster_rcnn_resnet50_lowproposals_coco_2018_01_28
- mask_rcnn_inception_resnet_v2_atrous_coco_2018_01_28
- mask_rcnn_inception_v2_coco_2018_01_28
- mask_rcnn_resnet101_atrous_coco_2018_01_28
- mask_rcnn_resnet50_atrous_coco_2018_01_28
- rfcn_resnet101_coco_2018_01_28
- ssd_inception_v2_coco_2018_01_28
- ssdlite_mobilenet_v2_coco_2018_05_09
- ssd_mobilenet_v1_0.75_depth_300x300_coco14_sync_2018_07_03
- ssd_mobilenet_v1_0.75_depth_quantized_300x300_coco14_sync_2018_07_18
- ssd_mobilenet_v1_coco_2018_01_28
- ssd_mobilenet_v1_fpn_shared_box_predictor_640x640_coco14_sync_2018_07_03
- ssd_mobilenet_v1_ppn_shared_box_predictor_300x300_coco14_sync_2018_07_03
- ssd_mobilenet_v1_quantized_300x300_coco14_sync_2018_07_18
- ssd_mobilenet_v2_coco_2018_03_29
- ssd_resnet50_v1_fpn_shared_box_predictor_640x640_coco14_sync_2018_07_03
I provide example ROS launch files for:
roslaunch object_detection_tensorflow ssd_mobilenet_v2.launch
roslaunch object_detection_tensorflow mask_rcnn_inception_v2.launch
- sensor_msgs/CompressedImage
- object_detection_tensorflow_msgs/BBoxArray
- ~camera_topic default: /image_raw - Input topic
- ~models_dir - directory where models are downloaded
- ~model_name - name of the directory where the model is located
- ~path_to_labels - path to labels.pbtxt
- ~num_classes default: 90 - number o classes
Node is an example of BBoxArray subscriber. It subscribes to BBoxArray and image, draws bboxes and publishes the result.
- sensor_msgs/CompressedImage
- object_detection_tensorflow_msgs/BBoxArray
- sensor_msgs/CompressedImage