Add EdgeTPU support (#3630)

* Add models/tf.py for TensorFlow and TFLite export

* Set auto=False for int8 calibration

* Update requirements.txt for TensorFlow and TFLite export

* Read anchors directly from PyTorch weights

* Add --tf-nms to append NMS in TensorFlow SavedModel and GraphDef export

* Remove check_anchor_order, check_file, set_logging from import

* Reformat code and optimize imports

* Autodownload model and check cfg

* update --source path, img-size to 320, single output

* Adjust representative_dataset

* Put representative dataset in tfl_int8 block

* detect.py TF inference

* weights to string

* weights to string

* cleanup tf.py

* Add --dynamic-batch-size

* Add xywh normalization to reduce calibration error

* Update requirements.txt

TensorFlow 2.3.1 -> 2.4.0 to avoid int8 quantization error

* Fix imports

Move C3 from models.experimental to models.common

* Add models/tf.py for TensorFlow and TFLite export

* Set auto=False for int8 calibration

* Update requirements.txt for TensorFlow and TFLite export

* Read anchors directly from PyTorch weights

* Add --tf-nms to append NMS in TensorFlow SavedModel and GraphDef export

* Remove check_anchor_order, check_file, set_logging from import

* Reformat code and optimize imports

* Autodownload model and check cfg

* update --source path, img-size to 320, single output

* Adjust representative_dataset

* detect.py TF inference

* Put representative dataset in tfl_int8 block

* weights to string

* weights to string

* cleanup tf.py

* Add --dynamic-batch-size

* Add xywh normalization to reduce calibration error

* Update requirements.txt

TensorFlow 2.3.1 -> 2.4.0 to avoid int8 quantization error

* Fix imports

Move C3 from models.experimental to models.common

* implement C3() and SiLU()

* Add TensorFlow and TFLite Detection

* Add --tfl-detect for TFLite Detection

* Add int8 quantized TFLite inference in detect.py

* Add --edgetpu for Edge TPU detection

* Fix --img-size to add rectangle TensorFlow and TFLite input

* Add --no-tf-nms to detect objects using models combined with TensorFlow NMS

* Fix --img-size list type input

* Update README.md

* Add Android project for TFLite inference

* Upgrade TensorFlow v2.3.1 -> v2.4.0

* Disable normalization of xywh

* Rewrite names init in detect.py

* Change input resolution 640 -> 320 on Android

* Disable NNAPI

* Update README.me --img 640 -> 320

* Update README.me for Edge TPU

* Update README.md

* Fix reshape dim to support dynamic batching

* Fix reshape dim to support dynamic batching

* Add epsilon argument in tf_BN, which is different between TF and PT

* Set stride to None if not using PyTorch, and do not warmup without PyTorch

* Add list support in check_img_size()

* Add list input support in detect.py

* sys.path.append('./') to run from yolov5/

* Add int8 quantization support for TensorFlow 2.5

* Add get_coco128.sh

* Remove --no-tfl-detect in models/tf.py (Use tf-android-tfl-detect branch for EdgeTPU)

* Update requirements.txt

* Replace torch.load() with attempt_load()

* Update requirements.txt

* Add --tf-raw-resize to set half_pixel_centers=False

* Remove android directory

* Update README.md

* Update README.md

* Add multiple OS support for EdgeTPU detection

* Fix export and detect

* Export 3 YOLO heads with Edge TPU models

* Remove xywh denormalization with Edge TPU models in detect.py

* Fix saved_model and pb detect error

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Fix pre-commit.ci failure

* Add edgetpu in export.py docstring

* Fix Edge TPU model detection exported by TF 2.7

* Add class names for TF/TFLite in DetectMultibackend

* Fix assignment with nl in TFLite Detection

* Add check when getting Edge TPU compiler version

* Add UTF-8 encoding in opening --data file for Windows

* Remove redundant TensorFlow import

* Add Edge TPU in export.py's docstring

* Add the detect layer in Edge TPU model conversion

* Default `dnn=False`

* Cleanup data.yaml loading

* Update detect.py

* Update val.py

* Comments and generalize data.yaml names

Co-authored-by: Glenn Jocher <glenn.jocher@ultralytics.com>
Co-authored-by: unknown <fangjiacong@ut.cn>
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>

detect.py

@@ -38,6 +38,7 @@ from utils.torch_utils import select_device, time_sync
 @torch.no_grad()
 def run(weights=ROOT / 'yolov5s.pt',  # model.pt path(s)
         source=ROOT / 'data/images',  # file/dir/URL/glob, 0 for webcam
+        data=ROOT / 'data/coco128.yaml',  # dataset.yaml path
         imgsz=(640, 640),  # inference size (height, width)
         conf_thres=0.25,  # confidence threshold
         iou_thres=0.45,  # NMS IOU threshold
@@ -76,7 +77,7 @@ def run(weights=ROOT / 'yolov5s.pt',  # model.pt path(s)
 
     # Load model
     device = select_device(device)
-    model = DetectMultiBackend(weights, device=device, dnn=dnn)
+    model = DetectMultiBackend(weights, device=device, dnn=dnn, data=data)
     stride, names, pt, jit, onnx, engine = model.stride, model.names, model.pt, model.jit, model.onnx, model.engine
     imgsz = check_img_size(imgsz, s=stride)  # check image size
 
@@ -204,6 +205,7 @@ def parse_opt():
     parser = argparse.ArgumentParser()
     parser.add_argument('--weights', nargs='+', type=str, default=ROOT / 'yolov5s.pt', help='model path(s)')
     parser.add_argument('--source', type=str, default=ROOT / 'data/images', help='file/dir/URL/glob, 0 for webcam')
+    parser.add_argument('--data', type=str, default=ROOT / 'data/coco128.yaml', help='(optional) dataset.yaml path')
     parser.add_argument('--imgsz', '--img', '--img-size', nargs='+', type=int, default=[640], help='inference size h,w')
     parser.add_argument('--conf-thres', type=float, default=0.25, help='confidence threshold')
     parser.add_argument('--iou-thres', type=float, default=0.45, help='NMS IoU threshold')

export.py

@@ -248,6 +248,24 @@ def export_tflite(keras_model, im, file, int8, data, ncalib, prefix=colorstr('Te
         LOGGER.info(f'\n{prefix} export failure: {e}')
 
 
+def export_edgetpu(keras_model, im, file, prefix=colorstr('Edge TPU:')):
+    # YOLOv5 Edge TPU export https://coral.ai/docs/edgetpu/models-intro/
+    try:
+        cmd = 'edgetpu_compiler --version'
+        out = subprocess.run(cmd, shell=True, capture_output=True, check=True)
+        ver = out.stdout.decode().split()[-1]
+        LOGGER.info(f'\n{prefix} starting export with Edge TPU compiler {ver}...')
+        f = str(file).replace('.pt', '-int8_edgetpu.tflite')
+        f_tfl = str(file).replace('.pt', '-int8.tflite')  # TFLite model
+
+        cmd = f"edgetpu_compiler -s {f_tfl}"
+        subprocess.run(cmd, shell=True, check=True)
+
+        LOGGER.info(f'{prefix} export success, saved as {f} ({file_size(f):.1f} MB)')
+    except Exception as e:
+        LOGGER.info(f'\n{prefix} export failure: {e}')
+
+
 def export_tfjs(keras_model, im, file, prefix=colorstr('TensorFlow.js:')):
     # YOLOv5 TensorFlow.js export
     try:
@@ -285,6 +303,7 @@ def export_tfjs(keras_model, im, file, prefix=colorstr('TensorFlow.js:')):
 
 
 def export_engine(model, im, file, train, half, simplify, workspace=4, verbose=False, prefix=colorstr('TensorRT:')):
+    # YOLOv5 TensorRT export https://developer.nvidia.com/tensorrt
     try:
         check_requirements(('tensorrt',))
         import tensorrt as trt
@@ -356,7 +375,7 @@ def run(data=ROOT / 'data/coco128.yaml',  # 'dataset.yaml path'
         ):
     t = time.time()
     include = [x.lower() for x in include]
-    tf_exports = list(x in include for x in ('saved_model', 'pb', 'tflite', 'tfjs'))  # TensorFlow exports
+    tf_exports = list(x in include for x in ('saved_model', 'pb', 'tflite', 'edgetpu', 'tfjs'))  # TensorFlow exports
     file = Path(url2file(weights) if str(weights).startswith(('http:/', 'https:/')) else weights)
 
     # Checks
@@ -405,15 +424,17 @@ def run(data=ROOT / 'data/coco128.yaml',  # 'dataset.yaml path'
 
     # TensorFlow Exports
     if any(tf_exports):
-        pb, tflite, tfjs = tf_exports[1:]
+        pb, tflite, edgetpu, tfjs = tf_exports[1:]
         assert not (tflite and tfjs), 'TFLite and TF.js models must be exported separately, please pass only one type.'
         model = export_saved_model(model, im, file, dynamic, tf_nms=nms or agnostic_nms or tfjs,
                                    agnostic_nms=agnostic_nms or tfjs, topk_per_class=topk_per_class, topk_all=topk_all,
                                    conf_thres=conf_thres, iou_thres=iou_thres)  # keras model
         if pb or tfjs:  # pb prerequisite to tfjs
             export_pb(model, im, file)
-        if tflite:
+        if tflite or edgetpu:
-            export_tflite(model, im, file, int8=int8, data=data, ncalib=100)
+            export_tflite(model, im, file, int8=int8 or edgetpu, data=data, ncalib=100)
+        if edgetpu:
+            export_edgetpu(model, im, file)
         if tfjs:
             export_tfjs(model, im, file)
 

models/common.py

@@ -17,6 +17,7 @@ import pandas as pd
 import requests
 import torch
 import torch.nn as nn
+import yaml
 from PIL import Image
 from torch.cuda import amp
 
@@ -276,7 +277,7 @@ class Concat(nn.Module):
 
 class DetectMultiBackend(nn.Module):
     # YOLOv5 MultiBackend class for python inference on various backends
-    def __init__(self, weights='yolov5s.pt', device=None, dnn=False):
+    def __init__(self, weights='yolov5s.pt', device=None, dnn=False, data=None):
         # Usage:
         #   PyTorch:      weights = *.pt
         #   TorchScript:            *.torchscript
@@ -284,6 +285,7 @@ class DetectMultiBackend(nn.Module):
         #   TensorFlow:             *_saved_model
         #   TensorFlow:             *.pb
         #   TensorFlow Lite:        *.tflite
+        #   TensorFlow Edge TPU:    *_edgetpu.tflite
         #   ONNX Runtime:           *.onnx
         #   OpenCV DNN:             *.onnx with dnn=True
         #   TensorRT:               *.engine
@@ -297,6 +299,9 @@ class DetectMultiBackend(nn.Module):
         pt, jit, onnx, engine, tflite, pb, saved_model, coreml = (suffix == x for x in suffixes)  # backend booleans
         stride, names = 64, [f'class{i}' for i in range(1000)]  # assign defaults
         w = attempt_download(w)  # download if not local
+        if data:  # data.yaml path (optional)
+            with open(data, errors='ignore') as f:
+                names = yaml.safe_load(f)['names']  # class names
 
         if jit:  # TorchScript
             LOGGER.info(f'Loading {w} for TorchScript inference...')
@@ -343,7 +348,7 @@ class DetectMultiBackend(nn.Module):
             binding_addrs = OrderedDict((n, d.ptr) for n, d in bindings.items())
             context = model.create_execution_context()
             batch_size = bindings['images'].shape[0]
-        else:  # TensorFlow model (TFLite, pb, saved_model)
+        else:  # TensorFlow (TFLite, pb, saved_model)
             if pb:  # https://www.tensorflow.org/guide/migrate#a_graphpb_or_graphpbtxt
                 LOGGER.info(f'Loading {w} for TensorFlow *.pb inference...')
                 import tensorflow as tf
@@ -425,6 +430,7 @@ class DetectMultiBackend(nn.Module):
             y[..., 1] *= h  # y
             y[..., 2] *= w  # w
             y[..., 3] *= h  # h
+
         y = torch.tensor(y) if isinstance(y, np.ndarray) else y
         return (y, []) if val else y
 

val.py

@@ -124,7 +124,7 @@ def run(data,
         (save_dir / 'labels' if save_txt else save_dir).mkdir(parents=True, exist_ok=True)  # make dir
 
         # Load model
-        model = DetectMultiBackend(weights, device=device, dnn=dnn)
+        model = DetectMultiBackend(weights, device=device, dnn=dnn, data=data)
         stride, pt, jit, engine = model.stride, model.pt, model.jit, model.engine
         imgsz = check_img_size(imgsz, s=stride)  # check image size
         half &= (pt or jit or engine) and device.type != 'cpu'  # half precision only supported by PyTorch on CUDA