Merge pull request #103 from Sidd1609:NanoDet

models/object_detection_nanodet/LICENSE

@@ -0,0 +1,202 @@
+
+                                 Apache License
+                           Version 2.0, January 2004
+                        http://www.apache.org/licenses/
+
+   TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
+
+   1. Definitions.
+
+      "License" shall mean the terms and conditions for use, reproduction,
+      and distribution as defined by Sections 1 through 9 of this document.
+
+      "Licensor" shall mean the copyright owner or entity authorized by
+      the copyright owner that is granting the License.
+
+      "Legal Entity" shall mean the union of the acting entity and all
+      other entities that control, are controlled by, or are under common
+      control with that entity. For the purposes of this definition,
+      "control" means (i) the power, direct or indirect, to cause the
+      direction or management of such entity, whether by contract or
+      otherwise, or (ii) ownership of fifty percent (50%) or more of the
+      outstanding shares, or (iii) beneficial ownership of such entity.
+
+      "You" (or "Your") shall mean an individual or Legal Entity
+      exercising permissions granted by this License.
+
+      "Source" form shall mean the preferred form for making modifications,
+      including but not limited to software source code, documentation
+      source, and configuration files.
+
+      "Object" form shall mean any form resulting from mechanical
+      transformation or translation of a Source form, including but
+      not limited to compiled object code, generated documentation,
+      and conversions to other media types.
+
+      "Work" shall mean the work of authorship, whether in Source or
+      Object form, made available under the License, as indicated by a
+      copyright notice that is included in or attached to the work
+      (an example is provided in the Appendix below).
+
+      "Derivative Works" shall mean any work, whether in Source or Object
+      form, that is based on (or derived from) the Work and for which the
+      editorial revisions, annotations, elaborations, or other modifications
+      represent, as a whole, an original work of authorship. For the purposes
+      of this License, Derivative Works shall not include works that remain
+      separable from, or merely link (or bind by name) to the interfaces of,
+      the Work and Derivative Works thereof.
+
+      "Contribution" shall mean any work of authorship, including
+      the original version of the Work and any modifications or additions
+      to that Work or Derivative Works thereof, that is intentionally
+      submitted to Licensor for inclusion in the Work by the copyright owner
+      or by an individual or Legal Entity authorized to submit on behalf of
+      the copyright owner. For the purposes of this definition, "submitted"
+      means any form of electronic, verbal, or written communication sent
+      to the Licensor or its representatives, including but not limited to
+      communication on electronic mailing lists, source code control systems,
+      and issue tracking systems that are managed by, or on behalf of, the
+      Licensor for the purpose of discussing and improving the Work, but
+      excluding communication that is conspicuously marked or otherwise
+      designated in writing by the copyright owner as "Not a Contribution."
+
+      "Contributor" shall mean Licensor and any individual or Legal Entity
+      on behalf of whom a Contribution has been received by Licensor and
+      subsequently incorporated within the Work.
+
+   2. Grant of Copyright License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      copyright license to reproduce, prepare Derivative Works of,
+      publicly display, publicly perform, sublicense, and distribute the
+      Work and such Derivative Works in Source or Object form.
+
+   3. Grant of Patent License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      (except as stated in this section) patent license to make, have made,
+      use, offer to sell, sell, import, and otherwise transfer the Work,
+      where such license applies only to those patent claims licensable
+      by such Contributor that are necessarily infringed by their
+      Contribution(s) alone or by combination of their Contribution(s)
+      with the Work to which such Contribution(s) was submitted. If You
+      institute patent litigation against any entity (including a
+      cross-claim or counterclaim in a lawsuit) alleging that the Work
+      or a Contribution incorporated within the Work constitutes direct
+      or contributory patent infringement, then any patent licenses
+      granted to You under this License for that Work shall terminate
+      as of the date such litigation is filed.
+
+   4. Redistribution. You may reproduce and distribute copies of the
+      Work or Derivative Works thereof in any medium, with or without
+      modifications, and in Source or Object form, provided that You
+      meet the following conditions:
+
+      (a) You must give any other recipients of the Work or
+          Derivative Works a copy of this License; and
+
+      (b) You must cause any modified files to carry prominent notices
+          stating that You changed the files; and
+
+      (c) You must retain, in the Source form of any Derivative Works
+          that You distribute, all copyright, patent, trademark, and
+          attribution notices from the Source form of the Work,
+          excluding those notices that do not pertain to any part of
+          the Derivative Works; and
+
+      (d) If the Work includes a "NOTICE" text file as part of its
+          distribution, then any Derivative Works that You distribute must
+          include a readable copy of the attribution notices contained
+          within such NOTICE file, excluding those notices that do not
+          pertain to any part of the Derivative Works, in at least one
+          of the following places: within a NOTICE text file distributed
+          as part of the Derivative Works; within the Source form or
+          documentation, if provided along with the Derivative Works; or,
+          within a display generated by the Derivative Works, if and
+          wherever such third-party notices normally appear. The contents
+          of the NOTICE file are for informational purposes only and
+          do not modify the License. You may add Your own attribution
+          notices within Derivative Works that You distribute, alongside
+          or as an addendum to the NOTICE text from the Work, provided
+          that such additional attribution notices cannot be construed
+          as modifying the License.
+
+      You may add Your own copyright statement to Your modifications and
+      may provide additional or different license terms and conditions
+      for use, reproduction, or distribution of Your modifications, or
+      for any such Derivative Works as a whole, provided Your use,
+      reproduction, and distribution of the Work otherwise complies with
+      the conditions stated in this License.
+
+   5. Submission of Contributions. Unless You explicitly state otherwise,
+      any Contribution intentionally submitted for inclusion in the Work
+      by You to the Licensor shall be under the terms and conditions of
+      this License, without any additional terms or conditions.
+      Notwithstanding the above, nothing herein shall supersede or modify
+      the terms of any separate license agreement you may have executed
+      with Licensor regarding such Contributions.
+
+   6. Trademarks. This License does not grant permission to use the trade
+      names, trademarks, service marks, or product names of the Licensor,
+      except as required for reasonable and customary use in describing the
+      origin of the Work and reproducing the content of the NOTICE file.
+
+   7. Disclaimer of Warranty. Unless required by applicable law or
+      agreed to in writing, Licensor provides the Work (and each
+      Contributor provides its Contributions) on an "AS IS" BASIS,
+      WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
+      implied, including, without limitation, any warranties or conditions
+      of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
+      PARTICULAR PURPOSE. You are solely responsible for determining the
+      appropriateness of using or redistributing the Work and assume any
+      risks associated with Your exercise of permissions under this License.
+
+   8. Limitation of Liability. In no event and under no legal theory,
+      whether in tort (including negligence), contract, or otherwise,
+      unless required by applicable law (such as deliberate and grossly
+      negligent acts) or agreed to in writing, shall any Contributor be
+      liable to You for damages, including any direct, indirect, special,
+      incidental, or consequential damages of any character arising as a
+      result of this License or out of the use or inability to use the
+      Work (including but not limited to damages for loss of goodwill,
+      work stoppage, computer failure or malfunction, or any and all
+      other commercial damages or losses), even if such Contributor
+      has been advised of the possibility of such damages.
+
+   9. Accepting Warranty or Additional Liability. While redistributing
+      the Work or Derivative Works thereof, You may choose to offer,
+      and charge a fee for, acceptance of support, warranty, indemnity,
+      or other liability obligations and/or rights consistent with this
+      License. However, in accepting such obligations, You may act only
+      on Your own behalf and on Your sole responsibility, not on behalf
+      of any other Contributor, and only if You agree to indemnify,
+      defend, and hold each Contributor harmless for any liability
+      incurred by, or claims asserted against, such Contributor by reason
+      of your accepting any such warranty or additional liability.
+
+   END OF TERMS AND CONDITIONS
+
+   APPENDIX: How to apply the Apache License to your work.
+
+      To apply the Apache License to your work, attach the following
+      boilerplate notice, with the fields enclosed by brackets "[]"
+      replaced with your own identifying information. (Don't include
+      the brackets!)  The text should be enclosed in the appropriate
+      comment syntax for the file format. We also recommend that a
+      file or class name and description of purpose be included on the
+      same "printed page" as the copyright notice for easier
+      identification within third-party archives.
+
+   Copyright [yyyy] [name of copyright owner]
+
+   Licensed under the Apache License, Version 2.0 (the "License");
+   you may not use this file except in compliance with the License.
+   You may obtain a copy of the License at
+
+       http://www.apache.org/licenses/LICENSE-2.0
+
+   Unless required by applicable law or agreed to in writing, software
+   distributed under the License is distributed on an "AS IS" BASIS,
+   WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+   See the License for the specific language governing permissions and
+   limitations under the License.

models/object_detection_nanodet/README.md

@@ -0,0 +1,119 @@
+# Nanodet
+
+Nanodet: NanoDet is a FCOS-style one-stage anchor-free object detection model which using Generalized Focal Loss as classification and regression loss.In NanoDet-Plus, we propose a novel label assignment strategy with a simple assign guidance module (AGM) and a dynamic soft label assigner (DSLA) to solve the optimal label assignment problem in lightweight model training.
+
+Note:
+- This version of nanodet: Nanodet-m-plus-1.5x_416
+
+## Demo
+
+Run the following command to try the demo: 
+```shell
+# detect on camera input
+python demo.py
+# detect on an image
+python demo.py --input /path/to/image
+```
+Note: 
+- image result saved as "result.jpg"
+
+
+## Results
+
+Here are some of the sample results that were observed using the model,
+
+![test1_res.jpg](./samples/1_res.jpg)
+![test2_res.jpg](./samples/2_res.jpg)
+  
+Video inference result,
+![WebCamR.gif](./samples/WebCamR.gif)
+
+## Model metrics:
+
+The model is evaluated on [COCO 2017 val](https://cocodataset.org/#download). Results are showed below:
+
+<table>
+<tr><th>Average Precision </th><th>Average Recall</th></tr>
+<tr><td>
+  
+|  area  |  IoU  |  Average Precision(AP)  |
+|:-------|:------|:------------------------|
+|  all  |  0.50:0.95  |  0.304  |
+|  all  |  0.50  |  0.459  |
+|  all  |  0.75  |  0.317  |
+|  small  |  0.50:0.95  |  0.107  |
+|  medium  |  0.50:0.95  |  0.322  |
+|  large  |  0.50:0.95  |  0.478  |
+ 
+ </td><td>
+
+  area  |  IoU  |  Average Recall  |
+|:-------|:------|:----------------|
+|  all  |  0.50:0.95  |  0.278  |
+|  all  |  0.50:0.95  |  0.434  |
+|  all  |  0.50:0.95 |  0.462  |
+|  small  |  0.50:0.95  |  0.198  |
+|  medium  |  0.50:0.95  |  0.510  |
+|  large  |  0.50:0.95  |  0.702  |
+</td></tr> </table>
+
+| class         | AP50   | mAP   | class          | AP50   | mAP   |
+|:--------------|:-------|:------|:---------------|:-------|:------|
+| person        | 67.5   | 41.8  | bicycle        | 35.4   | 18.8  |
+| car           | 45.0   | 25.4  | motorcycle     | 58.9   | 33.1  |
+| airplane      | 77.3   | 58.9  | bus            | 68.8   | 56.4  |
+| train         | 81.1   | 60.5  | truck          | 38.6   | 24.7  |
+| boat          | 35.5   | 16.7  | traffic light  | 30.5   | 14.0  |
+| fire hydrant  | 69.8   | 54.5  | stop sign      | 60.9   | 54.6  |
+| parking meter | 55.1   | 38.5  | bench          | 26.8   | 15.9  |
+| bird          | 38.3   | 23.6  | cat            | 82.5   | 62.1  |
+| dog           | 67.0   | 51.4  | horse          | 64.3   | 44.2  |
+| sheep         | 57.7   | 35.8  | cow            | 61.2   | 39.9  |
+| elephant      | 79.9   | 56.2  | bear           | 81.8   | 63.0  |
+| zebra         | 85.4   | 59.5  | giraffe        | 84.1   | 59.9  |
+| backpack      | 12.4   | 5.9   | umbrella       | 46.5   | 28.8  |
+| handbag       | 8.4    | 3.7   | tie            | 35.2   | 19.6  |
+| suitcase      | 38.1   | 23.8  | frisbee        | 60.7   | 43.9  |
+| skis          | 30.5   | 14.5  | snowboard      | 32.3   | 18.2  |
+| sports ball   | 37.6   | 24.5  | kite           | 51.1   | 30.4  |
+| baseball bat  | 28.9   | 13.6  | baseball glove | 40.1   | 21.6  |
+| skateboard    | 59.4   | 35.2  | surfboard      | 47.9   | 26.6  |
+| tennis racket | 55.2   | 30.5  | bottle         | 34.7   | 20.2  |
+| wine glass    | 27.8   | 16.3  | cup            | 35.5   | 23.7  |
+| fork          | 25.9   | 14.8  | knife          | 10.9   | 5.6   |
+| spoon         | 8.7    | 4.1   | bowl           | 42.8   | 29.4  |
+| banana        | 35.5   | 18.5  | apple          | 19.4   | 12.9  |
+| sandwich      | 46.7   | 33.4  | orange         | 35.2   | 25.9  |
+| broccoli      | 36.4   | 19.1  | carrot         | 30.9   | 17.8  |
+| hot dog       | 42.7   | 29.3  | pizza          | 61.0   | 44.9  |
+| donut         | 47.3   | 34.0  | cake           | 39.9   | 24.4  |
+| chair         | 28.8   | 16.1  | couch          | 60.5   | 42.6  |
+| potted plant  | 29.0   | 15.3  | bed            | 63.3   | 46.0  |
+| dining table  | 39.6   | 27.5  | toilet         | 71.3   | 55.3  |
+| tv            | 66.5   | 48.1  | laptop         | 62.6   | 46.9  |
+| mouse         | 63.5   | 44.1  | remote         | 19.8   | 10.3  |
+| keyboard      | 62.1   | 41.5  | cell phone     | 33.7   | 22.8  |
+| microwave     | 54.9   | 39.6  | oven           | 48.1   | 30.4  |
+| toaster       | 30.0   | 16.4  | sink           | 44.5   | 27.8  |
+| refrigerator  | 63.2   | 46.1  | book           | 18.4   | 7.3   |
+| clock         | 57.8   | 35.8  | vase           | 33.7   | 22.1  |
+| scissors      | 27.8   | 17.8  | teddy bear     | 54.1   | 35.4  |
+| hair drier    | 2.9    | 1.1   | toothbrush     | 13.1   | 8.2   |
+
+## License
+
+All files in this directory are licensed under [Apache 2.0 License](./LICENSE).
+
+#### Contributor Details
+
+- Google Summer of Code'22
+- Contributor: Sri Siddarth Chakaravarthy
+- Github Profile: https://github.com/Sidd1609
+- Organisation: OpenCV
+- Project: Lightweight object detection models using OpenCV 
+
+## Reference
+
+- Nanodet: https://zhuanlan.zhihu.com/p/306530300
+- Nanodet Plus: https://zhuanlan.zhihu.com/p/449912627
+- Nanodet weight and scripts for training: https://github.com/RangiLyu/nanodet

models/object_detection_nanodet/demo.py

@@ -0,0 +1,174 @@
+import numpy as np
+import cv2
+import argparse
+
+from nanodet import NanoDet
+
+def str2bool(v):
+    if v.lower() in ['on', 'yes', 'true', 'y', 't']:
+        return True
+    elif v.lower() in ['off', 'no', 'false', 'n', 'f']:
+        return False
+    else:
+        raise NotImplementedError
+
+backends = [cv2.dnn.DNN_BACKEND_OPENCV, cv2.dnn.DNN_BACKEND_CUDA]
+targets = [cv2.dnn.DNN_TARGET_CPU, cv2.dnn.DNN_TARGET_CUDA, cv2.dnn.DNN_TARGET_CUDA_FP16]
+help_msg_backends = "Choose one of the computation backends: {:d}: OpenCV implementation (default); {:d}: CUDA"
+help_msg_targets = "Chose one of the target computation devices: {:d}: CPU (default); {:d}: CUDA; {:d}: CUDA fp16"
+
+try:
+    backends += [cv2.dnn.DNN_BACKEND_TIMVX]
+    targets += [cv2.dnn.DNN_TARGET_NPU]
+    help_msg_backends += "; {:d}: TIMVX"
+    help_msg_targets += "; {:d}: NPU"
+except:
+    print('This version of OpenCV does not support TIM-VX and NPU. Visit https://github.com/opencv/opencv/wiki/TIM-VX-Backend-For-Running-OpenCV-On-NPU for more information.')
+
+classes = ('person', 'bicycle', 'car', 'motorcycle', 'airplane', 'bus',
+           'train', 'truck', 'boat', 'traffic light', 'fire hydrant',
+           'stop sign', 'parking meter', 'bench', 'bird', 'cat', 'dog',
+           'horse', 'sheep', 'cow', 'elephant', 'bear', 'zebra', 'giraffe',
+           'backpack', 'umbrella', 'handbag', 'tie', 'suitcase', 'frisbee',
+           'skis', 'snowboard', 'sports ball', 'kite', 'baseball bat',
+           'baseball glove', 'skateboard', 'surfboard', 'tennis racket',
+           'bottle', 'wine glass', 'cup', 'fork', 'knife', 'spoon', 'bowl',
+           'banana', 'apple', 'sandwich', 'orange', 'broccoli', 'carrot',
+           'hot dog', 'pizza', 'donut', 'cake', 'chair', 'couch',
+           'potted plant', 'bed', 'dining table', 'toilet', 'tv', 'laptop',
+           'mouse', 'remote', 'keyboard', 'cell phone', 'microwave',
+           'oven', 'toaster', 'sink', 'refrigerator', 'book', 'clock',
+           'vase', 'scissors', 'teddy bear', 'hair drier', 'toothbrush')
+
+def letterbox(srcimg, target_size=(416, 416)):
+    img = srcimg.copy()
+
+    top, left, newh, neww = 0, 0, target_size[0], target_size[1]
+    if img.shape[0] != img.shape[1]:
+        hw_scale = img.shape[0] / img.shape[1]
+        if hw_scale > 1:
+            newh, neww = target_size[0], int(target_size[1] / hw_scale)
+            img = cv2.resize(img, (neww, newh), interpolation=cv2.INTER_AREA)
+            left = int((target_size[1] - neww) * 0.5)
+            img = cv2.copyMakeBorder(img, 0, 0, left, target_size[1] - neww - left, cv2.BORDER_CONSTANT, value=0)  # add border
+        else:
+            newh, neww = int(target_size[0] * hw_scale), target_size[1]
+            img = cv2.resize(img, (neww, newh), interpolation=cv2.INTER_AREA)
+            top = int((target_size[0] - newh) * 0.5)
+            img = cv2.copyMakeBorder(img, top, target_size[0] - newh - top, 0, 0, cv2.BORDER_CONSTANT, value=0)
+    else:
+        img = cv2.resize(img, target_size, interpolation=cv2.INTER_AREA)
+
+    letterbox_scale = [top, left, newh, neww]
+    return img, letterbox_scale
+
+def unletterbox(bbox, original_image_shape, letterbox_scale):
+    ret = bbox.copy()
+
+    h, w = original_image_shape
+    top, left, newh, neww = letterbox_scale
+
+    if h == w:
+        ratio = h / newh
+        ret = ret * ratio
+        return ret
+
+    ratioh, ratiow = h / newh, w / neww
+    ret[0] = max((ret[0] - left) * ratiow, 0)
+    ret[1] = max((ret[1] - top) * ratioh, 0)
+    ret[2] = min((ret[2] - left) * ratiow, w)
+    ret[3] = min((ret[3] - top) * ratioh, h)
+
+    return ret.astype(np.int32)
+
+def vis(preds, res_img, letterbox_scale, fps=None):
+    ret = res_img.copy()
+
+    # draw FPS
+    if fps is not None:
+        fps_label = "FPS: %.2f" % fps
+        cv2.putText(ret, fps_label, (10, 25), cv2.FONT_HERSHEY_SIMPLEX, 1, (0, 0, 255), 2)
+
+    # draw bboxes and labels
+    for pred in preds:
+        bbox = pred[:4]
+        conf = pred[-2]
+        classid = pred[-1].astype(np.int32)
+
+        # bbox
+        xmin, ymin, xmax, ymax = unletterbox(bbox, ret.shape[:2], letterbox_scale)
+        cv2.rectangle(ret, (xmin, ymin), (xmax, ymax), (0, 255, 0), thickness=2)
+
+        # label
+        label = "{:s}: {:.2f}".format(classes[classid], conf)
+        cv2.putText(ret, label, (xmin, ymin - 10), cv2.FONT_HERSHEY_SIMPLEX, 1, (0, 255, 0), thickness=2)
+
+    return ret
+
+if __name__=='__main__':
+    parser = argparse.ArgumentParser(description='Nanodet inference using OpenCV an contribution by Sri Siddarth Chakaravarthy part of GSOC_2022')
+    parser.add_argument('--input', '-i', type=str, help='Path to the input image. Omit for using default camera.')
+    parser.add_argument('--model', '-m', type=str, default='object_detection_nanodet_2022nov.onnx', help="Path to the model")
+    parser.add_argument('--backend', '-b', type=int, default=backends[0], help=help_msg_backends.format(*backends))
+    parser.add_argument('--target', '-t', type=int, default=targets[0], help=help_msg_targets.format(*targets))
+    parser.add_argument('--confidence', default=0.35, type=float, help='Class confidence')
+    parser.add_argument('--nms', default=0.6, type=float, help='Enter nms IOU threshold')
+    parser.add_argument('--save', '-s', type=str2bool, default=False, help='Set true to save results. This flag is invalid when using camera.')
+    parser.add_argument('--vis', '-v', type=str2bool, default=True, help='Set true to open a window for result visualization. This flag is invalid when using camera.')
+    args = parser.parse_args()
+
+    model = NanoDet(modelPath= args.model,
+                    prob_threshold=args.confidence,
+                    iou_threshold=args.nms,
+                    backend_id=args.backend,
+                    target_id=args.target)
+
+    tm = cv2.TickMeter()
+    tm.reset()
+    if args.input is not None:
+        image = cv2.imread(args.input)
+        input_blob = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)
+
+        # Letterbox transformation
+        input_blob, letterbox_scale = letterbox(input_blob)
+
+        # Inference
+        tm.start()
+        preds = model.infer(input_blob)
+        tm.stop()
+        print("Inference time: {:.2f} ms".format(tm.getTimeMilli()))
+
+        img = vis(preds, image, letterbox_scale)
+
+        if args.save:
+            print('Resutls saved to result.jpg\n')
+            cv2.imwrite('result.jpg', img)
+
+        if args.vis:
+            cv2.namedWindow(args.input, cv2.WINDOW_AUTOSIZE)
+            cv2.imshow(args.input, img)
+            cv2.waitKey(0)
+
+    else:
+        print("Press any key to stop video capture")
+        deviceId = 0
+        cap = cv2.VideoCapture(deviceId)
+
+        while cv2.waitKey(1) < 0:
+            hasFrame, frame = cap.read()
+            if not hasFrame:
+                print('No frames grabbed!')
+                break
+
+            input_blob = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)
+            input_blob, letterbox_scale = letterbox(input_blob)
+            # Inference
+            tm.start()
+            preds = model.infer(input_blob)
+            tm.stop()
+
+            img = vis(preds, frame, letterbox_scale, fps=tm.getFPS())
+
+            cv2.imshow("NanoDet Demo", img)
+
+            tm.reset()

models/object_detection_nanodet/nanodet.py

@@ -0,0 +1,124 @@
+import numpy as np
+import cv2
+
+class NanoDet:
+    def __init__(self, modelPath, prob_threshold=0.35, iou_threshold=0.6, backend_id=0, target_id=0):
+        self.strides = (8, 16, 32, 64)
+        self.image_shape = (416, 416)
+        self.reg_max = 7
+        self.prob_threshold = prob_threshold
+        self.iou_threshold = iou_threshold
+        self.backend_id = backend_id
+        self.target_id = target_id
+        self.project = np.arange(self.reg_max + 1)
+        self.mean = np.array([103.53, 116.28, 123.675], dtype=np.float32).reshape(1, 1, 3)
+        self.std = np.array([57.375, 57.12, 58.395], dtype=np.float32).reshape(1, 1, 3)
+        self.net = cv2.dnn.readNet(modelPath)
+        self.net.setPreferableBackend(self.backend_id)
+        self.net.setPreferableTarget(self.target_id)
+
+        self.anchors_mlvl = []
+        for i in range(len(self.strides)):
+            featmap_size = (int(self.image_shape[0] / self.strides[i]), int(self.image_shape[1] / self.strides[i]))
+            stride = self.strides[i]
+            feat_h, feat_w = featmap_size
+            shift_x = np.arange(0, feat_w) * stride
+            shift_y = np.arange(0, feat_h) * stride
+            xv, yv = np.meshgrid(shift_x, shift_y)
+            xv = xv.flatten()
+            yv = yv.flatten()
+            cx = xv + 0.5 * (stride-1)
+            cy = yv + 0.5 * (stride - 1)
+            #anchors = np.stack((cx, cy), axis=-1)
+            anchors = np.column_stack((cx, cy))
+            self.anchors_mlvl.append(anchors)
+
+    @property
+    def name(self):
+        return self.__class__.__name__
+
+    def setBackend(self, backenId):
+        self.backend_id = backendId
+        self.net.setPreferableBackend(self.backend_id)
+
+    def setTarget(self, targetId):
+        self.target_id = targetId
+        self.net.setPreferableTarget(self.target_id)
+
+    def pre_process(self, img):
+        img = img.astype(np.float32)
+        img = (img - self.mean) / self.std
+        blob = cv2.dnn.blobFromImage(img)
+        return blob
+
+    def infer(self, srcimg):
+        blob = self.pre_process(srcimg)
+        self.net.setInput(blob)
+        outs = self.net.forward(self.net.getUnconnectedOutLayersNames())
+        preds = self.post_process(outs)
+        return preds
+
+    def post_process(self, preds):
+        cls_scores, bbox_preds = preds[::2], preds[1::2]
+        rescale = False
+        scale_factor = 1
+        bboxes_mlvl = []
+        scores_mlvl = []
+        for stride, cls_score, bbox_pred, anchors in zip(self.strides, cls_scores, bbox_preds, self.anchors_mlvl):
+            if cls_score.ndim==3:
+                cls_score = cls_score.squeeze(axis=0)
+            if bbox_pred.ndim==3:
+                bbox_pred = bbox_pred.squeeze(axis=0)
+
+            x_exp = np.exp(bbox_pred.reshape(-1, self.reg_max + 1))
+            x_sum = np.sum(x_exp, axis=1, keepdims=True)
+            bbox_pred = x_exp / x_sum
+            bbox_pred = np.dot(bbox_pred, self.project).reshape(-1,4)
+            bbox_pred *= stride
+
+            nms_pre = 1000
+            if nms_pre > 0 and cls_score.shape[0] > nms_pre:
+                max_scores = cls_score.max(axis=1)
+                topk_inds = max_scores.argsort()[::-1][0:nms_pre]
+                anchors = anchors[topk_inds, :]
+                bbox_pred = bbox_pred[topk_inds, :]
+                cls_score = cls_score[topk_inds, :]
+
+            points = anchors
+            distance = bbox_pred
+            max_shape=self.image_shape
+            x1 = points[:, 0] - distance[:, 0]
+            y1 = points[:, 1] - distance[:, 1]
+            x2 = points[:, 0] + distance[:, 2]
+            y2 = points[:, 1] + distance[:, 3]
+
+            if max_shape is not None:
+                x1 = np.clip(x1, 0, max_shape[1])
+                y1 = np.clip(y1, 0, max_shape[0])
+                x2 = np.clip(x2, 0, max_shape[1])
+                y2 = np.clip(y2, 0, max_shape[0])
+
+            #bboxes = np.stack([x1, y1, x2, y2], axis=-1)
+            bboxes = np.column_stack([x1, y1, x2, y2])
+            bboxes_mlvl.append(bboxes)
+            scores_mlvl.append(cls_score)
+
+        bboxes_mlvl = np.concatenate(bboxes_mlvl, axis=0)
+        if rescale:
+            bboxes_mlvl /= scale_factor
+        scores_mlvl = np.concatenate(scores_mlvl, axis=0)
+        bboxes_wh = bboxes_mlvl.copy()
+        bboxes_wh[:, 2:4] = bboxes_wh[:, 2:4] - bboxes_wh[:, 0:2]
+        classIds = np.argmax(scores_mlvl, axis=1)
+        confidences = np.max(scores_mlvl, axis=1)
+
+        indices = cv2.dnn.NMSBoxes(bboxes_wh.tolist(), confidences.tolist(), self.prob_threshold, self.iou_threshold)
+
+        if len(indices)>0:
+            det_bboxes = bboxes_mlvl[indices]
+            det_conf = confidences[indices]
+            det_classid = classIds[indices]
+
+            return np.concatenate([det_bboxes, det_conf.reshape(-1, 1), det_classid.reshape(-1, 1)], axis=1)
+        else:
+            return np.array([])