How to visualize prediction results

[ChenccLOL]

Hello, thanks for your great work.
I have used your code to complete the training and testing process. In the test results, I found that only the predicted mask results are stored. Could you please advise me how to convert the mask results into vector results in the ego vehicle coordinate system? Can you provide the corresponding visualization code? If possible, I would greatly appreciate it.

[RanRan814]

Hi! Have you ever visualized it? Thx

[GXmmm18]

Here is a simple code for visualization, hope it helps

import os
import numpy as np
import matplotlib.pyplot as plt
import tqdm
gt_paths = 'Path/to/your/data/nuscenes/customer/pivot-bezier/'
pred_paths = 'Path/to/your/evaluation/results/'
pred_npz_files = os.listdir(pred_paths)

def filter_and_denormalize(points, width, height):
    filtered_points = points[~np.all(points == 0, axis=1)]
    denormalized_points = filtered_points * [width, height]
    return denormalized_points

for f in tqdm.tqdm(pred_npz_files):
    prediction_data = np.load(pred_paths + f, allow_pickle=True)
    prediction_data = prediction_data['dt_res'].item()
    label_data  = np.load(gt_paths + f, allow_pickle=True)
    label_data = label_data['pivot_pts'].item()
    prediction_label_colors = {1: 'green', 2: 'gray', 3: 'red'}
    label_colors = {0: 'green', 1: 'gray', 2: 'red'}

    fig, axes = plt.subplots(1, 2, figsize=(8, 8))  

    ax = axes[0]
    points = prediction_data['map']
    labels = prediction_data['pred_label']
    for i, (pts, label) in enumerate(zip(points, labels)):
        if pts is not None:
            pts = np.array(pts)
            x, y = pts[:, 0], pts[:, 1]
            color = prediction_label_colors.get(label, 'blue')
            ax.plot(y, x, color=color)
            ax.scatter(y, x, color=color, s=10)

    ax.axis('off')
    ax.set_title('Predict', fontsize=14, color='blue')

    ax = axes[1]
    for label, arrays in label_data.items():
        for pts in arrays:
            if len(pts) > 0:
                pts = filter_and_denormalize(pts, 400, 200)
                if not np.all(pts == 0):  
                    x, y = pts[:, 0], pts[:, 1]
                    color = label_colors.get(label, 'blue')
                    ax.plot(y, x, color=color)
                    ax.scatter(y, x, color=color, s=10)

    ax.axis('off')
    ax.set_title('GroundTruth', fontsize=14, color='gray')


    plt.tight_layout()

    name = f.split('.')[0]
    plt.savefig(f'vis/{name}.png', bbox_inches='tight')
    plt.close()