models.py

from __future__ import division

import torch
import torch.nn as nn
import torch.nn.functional as F
from torch.autograd import Variable
import numpy as np

from utils.parse_config import *
from utils.utils import build_targets, to_cpu, non_max_suppression

import math
import config
from torch.distributions import Normal


def create_modules(module_defs, class_weight):
    """
    Constructs module list of layer blocks from module configuration in module_defs
    """
    hyperparams = module_defs.pop(0)
    output_filters = [int(hyperparams["channels"])]
    module_list = nn.ModuleList()
    for module_i, module_def in enumerate(module_defs):
        modules = nn.Sequential()

        if module_def["type"] == "convolutional":
            bn = int(module_def["batch_normalize"])
            filters = int(module_def["filters"])
            kernel_size = int(module_def["size"])
            pad = (kernel_size - 1) // 2
            modules.add_module(
                f"conv_{module_i}",
                nn.Conv2d(
                    in_channels=output_filters[-1],
                    out_channels=filters,
                    kernel_size=kernel_size,
                    stride=int(module_def["stride"]),
                    padding=pad,
                    bias=not bn,
                ),
            )
            if bn:
                modules.add_module(f"batch_norm_{module_i}", nn.BatchNorm2d(filters, momentum=0.9, eps=1e-5))
            if module_def["activation"] == "leaky":
                modules.add_module(f"leaky_{module_i}", nn.LeakyReLU(0.1))

        elif module_def["type"] == "maxpool":
            kernel_size = int(module_def["size"])
            stride = int(module_def["stride"])
            if kernel_size == 2 and stride == 1:
                modules.add_module(f"_debug_padding_{module_i}", nn.ZeroPad2d((0, 1, 0, 1)))
            maxpool = nn.MaxPool2d(kernel_size=kernel_size, stride=stride, padding=int((kernel_size - 1) // 2))
            modules.add_module(f"maxpool_{module_i}", maxpool)

        elif module_def["type"] == "upsample":
            upsample = Upsample(scale_factor=int(module_def["stride"]), mode="nearest")
            modules.add_module(f"upsample_{module_i}", upsample)

        elif module_def["type"] == "route":
            layers = [int(x) for x in module_def["layers"].split(",")]
            filters = sum([output_filters[1:][i] for i in layers])
            modules.add_module(f"route_{module_i}", EmptyLayer())

        elif module_def["type"] == "shortcut":
            filters = output_filters[1:][int(module_def["from"])]
            modules.add_module(f"shortcut_{module_i}", EmptyLayer())

        elif module_def["type"] == "yolo":
            anchor_idxs = [int(x) for x in module_def["mask"].split(",")]
            # Extract anchors
            anchors = [int(x) for x in module_def["anchors"].split(",")]
            anchors = [(anchors[i], anchors[i + 1]) for i in range(0, len(anchors), 2)]
            anchors = [anchors[i] for i in anchor_idxs]
            num_classes = int(module_def["classes"])
            img_size = int(hyperparams["height"])
            # Define detection layer
            yolo_layer = YOLOLayer(anchors, num_classes, img_size)
            modules.add_module(f"yolo_{module_i}", yolo_layer)
        elif module_def["type"] == "mdn":
            modules.add_module("mdn", MDN(output_filters[-1], int(module_def["filters"]), int(module_def["M"])))
        elif module_def["type"] == "fixed_mdn":
            modules.add_module("fixed_mdn", FixedMDN(
                        output_filters[-1],
                        int(module_def["filters"]),
                        int(module_def["M"]),
                        float(module_def["sigma"])))
        elif module_def["type"] == "hmdn":
            modules.add_module("hmdn", HMDN(
                        output_filters[-1],
                        int(module_def["filters"]),
                        int(module_def["M"]),
                        int(module_def["num_h"]),
                        float(module_def["eps"]),
                        class_weight))
 
            
        # Register module list and number of output filters
        module_list.append(modules)
        output_filters.append(filters)

    return hyperparams, module_list

class MDN(nn.Module):
    def __init__(self, in_channels, mid_channels, M):
        super(MDN, self).__init__()
        self.M = M
        self.net = nn.Sequential(
            nn.Conv2d(in_channels, M * 3, 3)
        )
        self.coef = 1.0 / math.sqrt(2*math.pi)

    def forward(self, data, score=None, clamp=None):
        x = self.net(data)
        x = x.view(data.shape[0], 3, self.M, -1)
        x = x.mean(-1)
        pi = nn.functional.softmax(x[:, 0, :], -1)
        mu = x[:, 1, :]
        #sigma = torch.exp(x[:, 2, :])
        #if clamp != None:
        #    sigma = sigma.clamp(0, clamp)
        sigma = torch.exp(x[:, 2, :])
        normal = Normal(mu, sigma)
        if score == None:
            return 0, pi, sigma, mu 
        else:
            log_probs = normal.log_prob(score.unsqueeze(1).expand_as(normal.loc))
            nll = -torch.logsumexp(torch.log(pi) + log_probs, dim=-1)
            nll = nll.mean()
            return nll, pi, sigma, mu

class FixedMDN(nn.Module):
    def __init__(self, in_channels, mid_channels, M, sigma):
        super(FixedMDN, self).__init__()
        self.M = M
        self.net = nn.Sequential(
            nn.Conv2d(in_channels, mid_channels, 3),
            nn.BatchNorm2d(mid_channels),
            nn.LeakyReLU(0.1),
            nn.Conv2d(mid_channels, M, 3)
        )
        self.mu = torch.arange(0, M, device=config.device).unsqueeze(0)
        self.sigma = torch.ones([1, M], device=config.device) * sigma

    def forward(self, data, score=None):
        x = self.net(data)
        x = x.view(data.shape[0], self.M, -1)
        x = x.mean(-1)
        log_pi = nn.functional.log_softmax(x, -1)
        mu = self.mu.expand_as(log_pi)
        sigma = self.sigma.expand_as(log_pi)
        normal = Normal(mu, sigma)
        if score == None:
            return 0, torch.exp(log_pi), sigma, mu 
        else:
            log_probs = normal.log_prob(score.unsqueeze(1).expand_as(normal.loc))
            nll = -torch.logsumexp(log_pi + log_probs, dim=-1)
            nll = nll.mean()
            return nll, torch.exp(log_pi), sigma, mu

class HMDN(nn.Module):
    def __init__(self, in_channels, mid_channels, M, num_h, eps, weight):
        super(HMDN, self).__init__()
        self.M = M
        self.num_h = num_h
        self.eps = eps
        self.hnet = nn.Sequential(
            nn.Conv2d(in_channels, mid_channels, 3),
            nn.BatchNorm2d(mid_channels),
            nn.LeakyReLU(0.1),
            nn.Conv2d(mid_channels, num_h, 3)
        )
        self.znet = nn.Sequential(
            nn.Linear(num_h, num_h * M * 2),
            nn.LeakyReLU(0.1),
            nn.Linear(num_h * M * 2, num_h * M),
        )
        self.avg_coef = torch.log(torch.tensor(1/num_h, device=config.device, requires_grad=False))
        self.weight = torch.tensor(weight, device=config.device, requires_grad=False) if weight is not None else None

    def forward(self, data, score=None):
        h = self.hnet(data)
        h = h.view(data.shape[0], self.num_h, -1)
        h = h.mean(-1)
        z = self.znet(h)
        z = z.view(-1, self.M, self.num_h)
        log_gamma = nn.functional.log_softmax(z, 1)
        log_pi = torch.logsumexp(self.avg_coef + log_gamma, dim=-1) 
        gamma = torch.exp(log_gamma)
        weight_sum = gamma.sum(-1) + 1e-4
        mu = (gamma * h.unsqueeze(1)).sum(-1) / weight_sum
        var = (gamma * (mu.unsqueeze(-1) - h.unsqueeze(1))**2).sum(-1) / weight_sum + 1e-4
        sigma = torch.sqrt(var)
        normal = Normal(mu, sigma)
        if score is None:
            return 0, 0, torch.exp(log_pi), sigma, mu, h
        else:
            dist = (h - score.unsqueeze(-1))**2
            min_dist, _ = torch.min(dist, -1)
            wta_loss = dist.mean(-1) * self.eps + min_dist * (1 - self.eps)
            wta_loss = wta_loss.mean()
            log_probs = normal.log_prob(score.unsqueeze(1).expand_as(normal.loc))
            nll = -torch.logsumexp(log_pi + log_probs, dim=-1)
            nll = self.weight[score.long()] * nll
            nll = nll.mean()
            return wta_loss, nll, torch.exp(log_pi), sigma, mu, h

class Upsample(nn.Module):
    """ nn.Upsample is deprecated """

    def __init__(self, scale_factor, mode="nearest"):
        super(Upsample, self).__init__()
        self.scale_factor = scale_factor
        self.mode = mode

    def forward(self, x):
        x = F.interpolate(x, scale_factor=self.scale_factor, mode=self.mode)
        return x


class EmptyLayer(nn.Module):
    """Placeholder for 'route' and 'shortcut' layers"""

    def __init__(self):
        super(EmptyLayer, self).__init__()


class YOLOLayer(nn.Module):
    """Detection layer"""

    def __init__(self, anchors, num_classes, img_dim=416):
        super(YOLOLayer, self).__init__()
        self.anchors = anchors
        self.num_anchors = len(anchors)
        self.num_classes = num_classes
        self.ignore_thres = 0.5
        self.mse_loss = nn.MSELoss()
        self.bce_loss = nn.BCELoss()
        self.obj_scale = 1
        self.noobj_scale = 100
        self.metrics = {}
        self.img_dim = img_dim
        self.grid_size = 0  # grid size

    def compute_grid_offsets(self, grid_size, cuda=True):
        self.grid_size = grid_size
        g = self.grid_size
        FloatTensor = torch.cuda.FloatTensor if cuda else torch.FloatTensor
        self.stride = self.img_dim / self.grid_size
        # Calculate offsets for each grid
        self.grid_x = torch.arange(g, device=config.device).repeat(g, 1).view([1, 1, g, g]).type(FloatTensor)
        self.grid_y = torch.arange(g, device=config.device).repeat(g, 1).t().view([1, 1, g, g]).type(FloatTensor)
        self.scaled_anchors = FloatTensor([(a_w / self.stride, a_h / self.stride) for a_w, a_h in self.anchors], device=config.device)
        self.anchor_w = self.scaled_anchors[:, 0:1].view((1, self.num_anchors, 1, 1))
        self.anchor_h = self.scaled_anchors[:, 1:2].view((1, self.num_anchors, 1, 1))

    def forward(self, x, targets=None, img_dim=None):

        # Tensors for cuda support
        FloatTensor = torch.cuda.FloatTensor if x.is_cuda else torch.FloatTensor
        LongTensor = torch.cuda.LongTensor if x.is_cuda else torch.LongTensor
        ByteTensor = torch.cuda.BoolTensor if x.is_cuda else torch.BoolTensor

        self.img_dim = img_dim
        num_samples = x.size(0)
        grid_size = x.size(2)

        prediction = (
            x.view(num_samples, self.num_anchors, self.num_classes + 5, grid_size, grid_size)
            .permute(0, 1, 3, 4, 2)
            .contiguous()
        )
        
        # Get outputs
        x = torch.sigmoid(prediction[..., 0])  # Center x
        y = torch.sigmoid(prediction[..., 1])  # Center y
        w = prediction[..., 2]  # Width
        h = prediction[..., 3]  # Height
        pred_conf = torch.sigmoid(prediction[..., 4])  # Conf
        pred_cls = torch.sigmoid(prediction[..., 5:])  # Cls pred.

        # If grid size does not match current we compute new offsets
        if grid_size != self.grid_size:
            self.compute_grid_offsets(grid_size, cuda=x.is_cuda)

        # Add offset and scale with anchors
        pred_boxes = FloatTensor(prediction[..., :4].shape, device=config.device)
        pred_boxes[..., 0] = x.data + self.grid_x
        pred_boxes[..., 1] = y.data + self.grid_y
        pred_boxes[..., 2] = torch.exp(w.data) * self.anchor_w
        pred_boxes[..., 3] = torch.exp(h.data) * self.anchor_h

        output = torch.cat(
            (
                pred_boxes.view(num_samples, -1, 4) * self.stride,
                pred_conf.view(num_samples, -1, 1),
                pred_cls.view(num_samples, -1, self.num_classes),
            ),
            -1,
        )

        if targets is None:
            return output, 0
        else:
            iou_scores, class_mask, obj_mask, noobj_mask, tx, ty, tw, th, tcls, tconf = build_targets(
                pred_boxes=pred_boxes,
                pred_cls=pred_cls,
                target=targets,
                anchors=self.scaled_anchors,
                ignore_thres=self.ignore_thres,
            )

            # Loss : Mask outputs to ignore non-existing objects (except with conf. loss)
            loss_x = self.mse_loss(x[obj_mask], tx[obj_mask])
            loss_y = self.mse_loss(y[obj_mask], ty[obj_mask])
            loss_w = self.mse_loss(w[obj_mask], tw[obj_mask])
            loss_h = self.mse_loss(h[obj_mask], th[obj_mask])
            loss_conf_obj = self.bce_loss(pred_conf[obj_mask], tconf[obj_mask])
            loss_conf_noobj = self.bce_loss(pred_conf[noobj_mask], tconf[noobj_mask])
            loss_conf = self.obj_scale * loss_conf_obj + self.noobj_scale * loss_conf_noobj
            loss_cls = self.bce_loss(pred_cls[obj_mask], tcls[obj_mask])
            total_loss = loss_x + loss_y + loss_w + loss_h + loss_conf + loss_cls

            # Metrics
            cls_acc = 100 * class_mask[obj_mask].mean()
            conf_obj = pred_conf[obj_mask].mean()
            conf_noobj = pred_conf[noobj_mask].mean()
            conf50 = (pred_conf > 0.5).float()
            iou50 = (iou_scores > 0.5).float()
            iou75 = (iou_scores > 0.75).float()
            detected_mask = conf50 * class_mask * tconf
            precision = torch.sum(iou50 * detected_mask) / (conf50.sum() + 1e-16)
            recall50 = torch.sum(iou50 * detected_mask) / (obj_mask.sum() + 1e-16)
            recall75 = torch.sum(iou75 * detected_mask) / (obj_mask.sum() + 1e-16)

            self.metrics = {
                "loss": to_cpu(total_loss).item(),
                "x": to_cpu(loss_x).item(),
                "y": to_cpu(loss_y).item(),
                "w": to_cpu(loss_w).item(),
                "h": to_cpu(loss_h).item(),
                "conf": to_cpu(loss_conf).item(),
                "cls": to_cpu(loss_cls).item(),
                "cls_acc": to_cpu(cls_acc).item(),
                "recall50": to_cpu(recall50).item(),
                "recall75": to_cpu(recall75).item(),
                "precision": to_cpu(precision).item(),
                "conf_obj": to_cpu(conf_obj).item(),
                "conf_noobj": to_cpu(conf_noobj).item(),
                "grid_size": grid_size,
            }

            return output, total_loss


class Darknet(nn.Module):
    """YOLOv3 object detection model"""

    def __init__(self, module_defs, class_weight=None, img_size=416):
        super(Darknet, self).__init__()
        self.module_defs = module_defs
        self.hyperparams, self.module_list = create_modules(self.module_defs, class_weight)
        self.yolo_layers = [layer[0] for layer in self.module_list if hasattr(layer[0], "metrics")]
        self.img_size = img_size
        self.seen = 0
        self.header_info = np.array([0, 0, 0, self.seen, 0], dtype=np.int32)

    def forward(self, x, targets=None, scores=None):
        img_dim = x.shape[2]
        loss = 0
        layer_outputs, yolo_outputs = [], []
        mdn_output = None
        for i, (module_def, module) in enumerate(zip(self.module_defs, self.module_list)):
            if module_def["type"] in ["convolutional", "upsample", "maxpool"]:
                x = module(x)
            elif module_def["type"] == "route":
                x = torch.cat([layer_outputs[int(layer_i)] for layer_i in module_def["layers"].split(",")], 1)
            elif module_def["type"] == "shortcut":
                layer_i = int(module_def["from"])
                x = layer_outputs[-1] + layer_outputs[layer_i]
            elif module_def["type"] == "yolo":
                x, layer_loss = module[0](x, targets, img_dim)
                loss += layer_loss
                yolo_outputs.append(x)
            elif module_def["type"] == "mdn":
                mdn_output = module[0](x, scores, clamp)
            elif module_def["type"] == "fixed_mdn":
                mdn_output = module[0](x, scores)
            elif module_def["type"] == "hmdn":
                mdn_output = module[0](x, scores)
            layer_outputs.append(x)
        
        if len(yolo_outputs) != 0:
            yolo_outputs = to_cpu(torch.cat(yolo_outputs, 1))
        if mdn_output is not None:
            mdn_output = [mdn_output[0], mdn_output[1]] + [to_cpu(o) for o in mdn_output[2:]]

        if targets is not None:
            if mdn_output is not None:
                return loss, yolo_outputs, mdn_output
            else:
                return loss, yolo_outputs
        else:
            if mdn_output is not None:
                return yolo_outputs, mdn_output
            else:
                return yolo_outputs

    def load_darknet_weights(self, weights_path):
        """Parses and loads the weights stored in 'weights_path'"""

        # Open the weights file
        with open(weights_path, "rb") as f:
            header = np.fromfile(f, dtype=np.int32, count=5)  # First five are header values
            self.header_info = header  # Needed to write header when saving weights
            self.seen = header[3]  # number of images seen during training
            weights = np.fromfile(f, dtype=np.float32)  # The rest are weights

        # Establish cutoff for loading backbone weights
        cutoff = None
        if "darknet53.conv.74" in weights_path:
            cutoff = 75
        elif "yolov3-tiny" in weights_path:
            #temporarily
            cutoff = 10

        ptr = 0
        for i, (module_def, module) in enumerate(zip(self.module_defs, self.module_list)):
            if i == cutoff:
                break
            if module_def["type"] == "convolutional":
                conv_layer = module[0]
                if module_def["batch_normalize"]:
                    # Load BN bias, weights, running mean and running variance
                    bn_layer = module[1]
                    num_b = bn_layer.bias.numel()  # Number of biases
                    # Bias
                    bn_b = torch.from_numpy(weights[ptr : ptr + num_b]).view_as(bn_layer.bias)
                    bn_layer.bias.data.copy_(bn_b)
                    ptr += num_b
                    # Weight
                    bn_w = torch.from_numpy(weights[ptr : ptr + num_b]).view_as(bn_layer.weight)
                    bn_layer.weight.data.copy_(bn_w)
                    ptr += num_b
                    # Running Mean
                    bn_rm = torch.from_numpy(weights[ptr : ptr + num_b]).view_as(bn_layer.running_mean)
                    bn_layer.running_mean.data.copy_(bn_rm)
                    ptr += num_b
                    # Running Var
                    bn_rv = torch.from_numpy(weights[ptr : ptr + num_b]).view_as(bn_layer.running_var)
                    bn_layer.running_var.data.copy_(bn_rv)
                    ptr += num_b
                else:
                    # Load conv. bias
                    num_b = conv_layer.bias.numel()
                    conv_b = torch.from_numpy(weights[ptr : ptr + num_b]).view_as(conv_layer.bias)
                    conv_layer.bias.data.copy_(conv_b)
                    ptr += num_b
                # Load conv. weights
                num_w = conv_layer.weight.numel()
                conv_w = torch.from_numpy(weights[ptr : ptr + num_w]).view_as(conv_layer.weight)
                conv_layer.weight.data.copy_(conv_w)
                ptr += num_w

    def save_darknet_weights(self, path, cutoff=-1):
        """
            @:param path    - path of the new weights file
            @:param cutoff  - save layers between 0 and cutoff (cutoff = -1 -> all are saved)
        """
        fp = open(path, "wb")
        self.header_info[3] = self.seen
        self.header_info.tofile(fp)

        # Iterate through layers
        for i, (module_def, module) in enumerate(zip(self.module_defs[:cutoff], self.module_list[:cutoff])):
            if module_def["type"] == "convolutional":
                conv_layer = module[0]
                # If batch norm, load bn first
                if module_def["batch_normalize"]:
                    bn_layer = module[1]
                    bn_layer.bias.data.cpu().numpy().tofile(fp)
                    bn_layer.weight.data.cpu().numpy().tofile(fp)
                    bn_layer.running_mean.data.cpu().numpy().tofile(fp)
                    bn_layer.running_var.data.cpu().numpy().tofile(fp)
                # Load conv bias
                else:
                    conv_layer.bias.data.cpu().numpy().tofile(fp)
                # Load conv weights
                conv_layer.weight.data.cpu().numpy().tofile(fp)

        fp.close()