Unable to perform inference on pretrained weights

[AditTuf]

Hi Aladdin great tutorials you have here. I was really able to understand for the first time how to code YOLOv3 . But i couldn't find the code for inference so i decided to write one on my own but i stumbled across following issues.

1. I tried to use your plot_couple_examples function ,but always got cuda out of memory error ,which was really weird cause i could train the model for batch size 6 on my 4gb gtx1650 gpu
2. Then i tried to manually pass an image by reading through opencv and then expanding dimension to have batch size of 1 but got the following error
3. RuntimeError: Given groups=1, weight of size [32, 3, 3, 3], expected input[1, 416, 416, 3] to have 3 channels, but got 416 channels instead
4. Can you please suggest some fix i am using your repo as starting point for my college project

[AditTuf]

I have written the inference code myself but besides the very first image in PASCAL_VOC where it is able to detect a dog , there is no detection for any object for rest of many images.

After printing some outputs i could see that many spurious bounding boxes were detected , but no box actually came through nms stage and hence no detection on very easy examples of person or dogs .

Here is code i've written

import config
import torch
import torch.optim as optim

from model import YOLOv3
from tqdm import tqdm
import cv2
from loss import YoloLoss
import warnings
warnings.filterwarnings("ignore")
from utils import (
    mean_average_precision,
    load_checkpoint,get_loaders, cells_to_bboxes,non_max_suppression,plot_image
)
torch.backends.cudnn.benchmark = True
import numpy as np

def plot_couple_examples(model,thresh, iou_thresh, anchors):
    model.eval()
    anchors = torch.tensor(anchors)
    anchors = anchors.to(config.DEVICE)
    x = cv2.imread('PASCAL_VOC/images/001931.jpg')#,cv2.COLOR_BGR2RGB 000001.jpg
    x = cv2.resize(x,(416,416))
    x= np.rollaxis(x, 2, 0)
    x = np.expand_dims(x,0)  
    x = torch.tensor(x)
    x = x.type(torch.cuda.FloatTensor)
    x = x.to("cuda")
    with torch.no_grad():
        out = model(x)
        bboxes = [[] for _ in range(x.shape[0])]
        for i in range(1):
            batch_size, A, S, _, _ = out[i].shape
            anchor = anchors[i]
            boxes_scale_i = cells_to_bboxes(
                out[i], anchor, S=S, is_preds=True
            )
            for idx, (box) in enumerate(boxes_scale_i):
                bboxes[idx] += box

        model.train()

    for i in range(batch_size):
        print(bboxes)
        nms_boxes = non_max_suppression(
            bboxes[i], iou_threshold=iou_thresh, threshold=thresh, box_format="midpoint",
        )
        print(nms_boxes)
        x = cv2.imread('PASCAL_VOC/images/001931.jpg',cv2.COLOR_BGR2RGB)
        x = cv2.resize(x,(416,416))
        plot_image(x, nms_boxes)

model = YOLOv3(num_classes=config.NUM_CLASSES).to(config.DEVICE)

optimizer = optim.Adam(
        model.parameters(), lr=config.LEARNING_RATE, weight_decay=config.WEIGHT_DECAY
    )
load_checkpoint(
        config.CHECKPOINT_FILE, model, optimizer, config.LEARNING_RATE
    )
anchors = config.ANCHORS
plot_couple_examples(model,0.3, 0.3, anchors)

I am again reading the image because the original code was modifying the image in some odd ways

[AditTuf]

Now i am getting some detections using the following code for inference ,however bounding box height and width seem to be very small .Plus I was getting no detections when i used my previous code in above comment can you please tell some solution to this.

import config
import torch
import torch.optim as optim

from model import YOLOv3
from tqdm import tqdm
import cv2
from loss import YoloLoss
import warnings
warnings.filterwarnings("ignore")
from utils import (
    mean_average_precision,
    load_checkpoint,get_loaders, cells_to_bboxes,non_max_suppression,plot_image,plot_couple_examples
)
torch.backends.cudnn.benchmark = True
import numpy as np

model = YOLOv3(num_classes=config.NUM_CLASSES).to(config.DEVICE)

optimizer = optim.Adam(
        model.parameters(), lr=config.LEARNING_RATE, weight_decay=config.WEIGHT_DECAY
    )
load_checkpoint(
        config.CHECKPOINT_FILE, model, optimizer, config.LEARNING_RATE
    )
train_loader, test_loader, train_eval_loader = get_loaders(
        train_csv_path=config.DATASET + "/8examples.csv", test_csv_path=config.DATASET + "/8examples.csv"
    )

anchors = config.ANCHORS
anchors = torch.tensor(anchors)
anchors = anchors.to(config.DEVICE)
plot_couple_examples(model,test_loader,0.85, 0.85, anchors)		

[SimoDB90]

hi! Have you found a solution?
I can't make it work