EfficientNMS.py

import torch
import torch.nn as nn

class TRT_NMS(torch.autograd.Function):

    @staticmethod
    def forward(
        ctx,
        boxes,
        scores,
        background_class=-1,
        box_coding=0,
        iou_threshold=0.45,
        max_output_boxes=100,
        plugin_version="1",
        score_activation=0,
        score_threshold=0.25
    ):
        batch_size, num_boxes, num_classes = scores.shape
        num_det = torch.randint(0, max_output_boxes, (batch_size, 1),dtype=torch.int32)
        det_boxes = torch.randn(batch_size, max_output_boxes, 4)
        det_scores = torch.randn(batch_size, max_output_boxes)
        det_classes = torch.randint(0, num_classes, (batch_size, max_output_boxes),dtype=torch.int32)

        return num_det, det_boxes, det_scores, det_classes

    @staticmethod
    def symbolic(g,
                 boxes,
                 scores,
                 background_class=-1,
                 box_coding=0,
                 iou_threshold=0.45,
                 max_output_boxes=100,
                 plugin_version="1",
                 score_activation=0,
                 score_threshold=0.25):
        out = g.op("TRT::EfficientNMS_TRT",
                   boxes,
                   scores,
                   background_class_i=background_class,
                   box_coding_i=box_coding,
                   iou_threshold_f=iou_threshold,
                   max_output_boxes_i=max_output_boxes,
                   plugin_version_s=plugin_version,
                   score_activation_i=score_activation,
                   score_threshold_f=score_threshold,
                   outputs=4)
        nums, boxes, scores, classes = out
        return nums, boxes, scores, classes

class ONNX_TRT(nn.Module):
    def __init__(self, max_obj=100, iou_thres=0.45, score_thres=0.25, max_wh=False, device=None):
        super().__init__()
        assert max_wh is False
        self.device = device
        self.background_class = -1,
        self.box_coding = 0,
        self.iou_threshold = iou_thres
        self.max_obj = max_obj
        self.plugin_version = '1'
        self.score_activation = 0
        self.score_threshold = score_thres
        self.convert_matrix = torch.tensor([[1, 0, 1, 0], [0, 1, 0, 1], [-0.5, 0, 0.5, 0], [0, -0.5, 0, 0.5]],
                                           dtype=torch.float32,device=self.device)
    def forward(self, x):
        box = x[:, :, :4]
        conf = x[:, :, 4:5]
        score = x[:, :, 5:]
        score *= conf
        box @= self.convert_matrix
        num_det, det_boxes, det_scores, det_classes = TRT_NMS.apply(box, score, self.background_class, self.box_coding,
                                                                    self.iou_threshold, self.max_obj,
                                                                    self.plugin_version, self.score_activation,
                                                                    self.score_threshold)
        return num_det, det_boxes, det_scores, det_classes


class End2End(nn.Module):

    def __init__(self, model, max_obj=100, iou_thres=0.45, score_thres=0.25, max_wh=None, device=None):
        super().__init__()
        device = device
        self.model = model.to(device)
        self.patch_model = ONNX_TRT
        self.end2end = self.patch_model(max_obj, iou_thres, score_thres, max_wh, device)
        self.end2end.eval()

    def forward(self, x):
        x = x[:,[2,1,0],...]
        x = x * (1/255)
        x = self.model(x)[0]
        x = self.end2end(x)
        return x