def inicializar_pointnet(): print("Loading 3D Object Classification Model:") start = time.time() global classifier classifier = PointNetCls(k=2) #IMP, 2 é o número de classes, depois alterar. classifier.cuda() #classifier.load_state_dict(torch.load('/home/socialab/FCHarDNet/cls_model_40.pth')) classifier.load_state_dict(torch.load('/home/socialab/human_vision/FCHarDNet/cls_model_40.pth')) classifier.eval() end = time.time() print(" (time): " + str(end-start))
loss += feature_transform_regularizer(trans_feat) * 0.001 loss.backward() optimizer.step() pred_choice = pred.data.max(1)[1] correct = pred_choice.eq(target.data).cpu().sum() print('[%d: %d/%d] train loss: %f accuracy: %f' % (epoch, i, num_batch, loss.item(), correct.item() / float(opt.batchSize))) if i % 10 == 0: j, data = next(enumerate(testdataloader, 0)) points, target = data target = target[:, 0] points = points.transpose(2, 1) points, target = points.cuda(), target.cuda() classifier = classifier.eval() pred, _, _ = classifier(points) loss = F.nll_loss(pred, target) pred_choice = pred.data.max(1)[1] correct = pred_choice.eq(target.data).cpu().sum() print('[%d: %d/%d] %s loss: %f accuracy: %f' % (epoch, i, num_batch, blue('test'), loss.item(), correct.item() / float(opt.batchSize))) torch.save(classifier.state_dict(), '%s/cls_model_%d.pth' % (opt.outf, epoch)) total_correct = 0 total_testset = 0 for i, data in tqdm(enumerate(testdataloader, 0)): points, target = data
os.makedirs(os.path.join('save', model_name, adv+'-'+str(eps)+'-'+str(int(n))+'-'+str(eps_iter))) with open(os.path.join('dataset', 'random1024', 'whole_data_and_whole_label.pkl'), 'rb') as fid: whole_data, whole_label = pkl.load(fid) if model_name == 'PointNet': from pointnet.model import PointNetCls model = PointNetCls(k=40, feature_transform=True, predict_logit=True) checkpoint = 'pointnet/cls_model_201.pth' else: print('No such model architecture') assert False model = model.to(device) model.load_state_dict(torch.load(checkpoint)) model.eval() pytorch_utils.requires_grad_(model, False) print("Model name\t%s" % model_name) cnt = 0 # adv pointcloud successfully attacked CNT = 0 # clean pointcloud correctly classified for idx in tqdm(range(len(whole_data))): x = whole_data[idx] label = whole_label[idx] with torch.no_grad(): y_pred = model(torch.from_numpy(x[np.newaxis,:,:]).float().to(device)) y_pred_idx = np.argmax(y_pred.detach().cpu().numpy().flatten())
optimizer.step() # 梯度下降,参数优化 pred_choice = pred.data.max(1)[1] # max(1)返回每一行中的最大值及索引,[1]取出索引(代表着类别) correct = pred_choice.eq( target.data).cpu().sum() # 判断和target是否匹配,并计算匹配的数量 print('[%d: %d/%d] train loss: %f accuracy: %f' % (epoch, i, num_batch, loss.item(), correct.item() / float(opt.batchSize))) # 每10次batch之后,进行一次测试 if i % 10 == 0: j, data = next(enumerate(testdataloader, 0)) points, target = data target = target[:, 0] points = points.transpose(2, 1) points, target = points.cuda(), target.cuda() classifier = classifier.eval() # 测试模式,固定住BN和dropout pred, _, _ = classifier(points) loss = F.nll_loss(pred, target) pred_choice = pred.data.max(1)[1] correct = pred_choice.eq(target.data).cpu().sum() print('[%d: %d/%d] %s loss: %f accuracy: %f' % (epoch, i, num_batch, blue('test'), loss.item(), correct.item() / float(opt.batchSize))) torch.save(classifier.state_dict(), '%s/cls_model_%d.pth' % (opt.outf, epoch)) total_correct = 0 total_testset = 0 for i, data in tqdm(enumerate(testdataloader, 0)): points, target = data
end_time = time.time() print("classification~ cost:{:.1f}ms ".format( (end_time - start_time) * 1000)) print('-------------------------------------') def listener(): # In ROS, nodes are uniquely named. If two nodes with the same # name are launched, the previous one is kicked off. The # anonymous=True flag means that rospy will choose a unique # name for our 'listener' node so that multiple listeners can # run simultaneously. rospy.init_node('obj_classify', anonymous=True) rospy.Subscriber("fused_lidar", pc2, callback) global pub pub = rospy.Publisher("classification", pc2, queue_size=20) # spin() simply keeps python from exiting until this node is stopped rospy.spin() if __name__ == '__main__': # 网络加载 classifier = PointNetCls(k=num_classes) # 训练模型为16类 classifier.cuda() classifier.load_state_dict(torch.load(path_model)) classifier.eval() # 测试模式 pub = None listener()
def our_main(): from utils.show3d_balls import showpoints parser = argparse.ArgumentParser() parser.add_argument( '--batchSize', type=int, default=32, help='input batch size') parser.add_argument( '--num_points', type=int, default=2000, help='input batch size') parser.add_argument( '--workers', type=int, help='number of data loading workers', default=4) parser.add_argument( '--nepoch', type=int, default=250, help='number of epochs to train for') parser.add_argument('--outf', type=str, default='cls', help='output folder') parser.add_argument('--model', type=str, default='', help='model path') parser.add_argument('--dataset', type=str, required=True, help="dataset path") parser.add_argument('--dataset_type', type=str, default='shapenet', help="dataset type shapenet|modelnet40") parser.add_argument('--feature_transform', action='store_true', help="use feature transform") opt = parser.parse_args() print(opt) blue = lambda x: '\033[94m' + x + '\033[0m' opt.manualSeed = random.randint(1, 10000) # fix seed print("Random Seed: ", opt.manualSeed) random.seed(opt.manualSeed) torch.manual_seed(opt.manualSeed) if opt.dataset_type == 'shapenet': dataset = ShapeNetDataset( root=opt.dataset, classification=True, npoints=opt.num_points) test_dataset = ShapeNetDataset( root=opt.dataset, classification=True, split='test', npoints=opt.num_points, data_augmentation=False) elif opt.dataset_type == 'modelnet40': dataset = ModelNetDataset( root=opt.dataset, npoints=opt.num_points, split='trainval') test_dataset = ModelNetDataset( root=opt.dataset, split='test', npoints=opt.num_points, data_augmentation=False) else: exit('wrong dataset type') dataloader = torch.utils.data.DataLoader( dataset, batch_size=opt.batchSize, shuffle=True, num_workers=int(opt.workers)) testdataloader = torch.utils.data.DataLoader( test_dataset, batch_size=opt.batchSize, shuffle=True, num_workers=int(opt.workers)) print(len(dataset), len(test_dataset)) num_classes = len(dataset.classes) print('classes', num_classes) try: os.makedirs(opt.outf) except OSError: pass classifier = PointNetCls(k=num_classes, feature_transform=opt.feature_transform) if opt.model != '': classifier.load_state_dict(torch.load(opt.model)) optimizer = optim.Adam(classifier.parameters(), lr=0.001, betas=(0.9, 0.999)) scheduler = optim.lr_scheduler.StepLR(optimizer, step_size=20, gamma=0.5) classifier.cuda() num_batch = len(dataset) / opt.batchSize ## python train_classification.py --dataset ../dataset --nepoch=4 --dataset_type shapenet for epoch in range(opt.nepoch): scheduler.step() for i, data in enumerate(dataloader, 0): points, target = data target = target[:, 0] showpoints(points[0].numpy()) points = points.transpose(2, 1) points, target = points.cuda(), target.cuda() optimizer.zero_grad() classifier = classifier.train() pred, trans, trans_feat = classifier(points) loss = F.nll_loss(pred, target) if opt.feature_transform: loss += feature_transform_regularizer(trans_feat) * 0.001 loss.backward() optimizer.step() pred_choice = pred.data.max(1)[1] correct = pred_choice.eq(target.data).cpu().sum() print('[%d: %d/%d] train loss: %f accuracy: %f' % (epoch, i, num_batch, loss.item(), correct.item() / float(opt.batchSize))) if i % 10 == 0: j, data = next(enumerate(testdataloader, 0)) points, target = data target = target[:, 0] points = points.transpose(2, 1) points, target = points.cuda(), target.cuda() classifier = classifier.eval() pred, _, _ = classifier(points) loss = F.nll_loss(pred, target) pred_choice = pred.data.max(1)[1] correct = pred_choice.eq(target.data).cpu().sum() print('[%d: %d/%d] %s loss: %f accuracy: %f' % (epoch, i, num_batch, blue('test'), loss.item(), correct.item()/float(opt.batchSize))) torch.save(classifier.state_dict(), '%s/cls_model_%d.pth' % (opt.outf, epoch)) total_correct = 0 total_testset = 0 for i,data in tqdm(enumerate(testdataloader, 0)): points, target = data target = target[:, 0] points = points.transpose(2, 1) points, target = points.cuda(), target.cuda() classifier = classifier.eval() pred, _, _ = classifier(points) pred_choice = pred.data.max(1)[1] correct = pred_choice.eq(target.data).cpu().sum() total_correct += correct.item() total_testset += points.size()[0] print("final accuracy {}".format(total_correct / float(total_testset)))
if opt.feature_transform: loss += feature_transform_regularizer(trans_feat) * 0.001 loss.backward() # 小批量的损失对模型参数求梯度 optimizer.step() # 梯度下降优化 以batch为单位, 通过调用optim实例的step函数来迭代模型参数, w, b pred_choice = pred.data.max(1)[1] correct = pred_choice.eq(target.data).cpu().sum().item() # n += target.shape[0] print('[%d: %d/%d] train loss: %f accuracy: %f' % (epoch, i, num_batch, loss.item(), correct / float(opt.batchSize))) if i % 10 == 0: # 每10个batch执行一次,即为每10*batchSize个点云执行一次验证 j, data = next(enumerate(testdataloader, 0)) points, target = data target = target[:, 0] points = points.transpose(2, 1) points, target = points.cuda(), target.cuda() # 复制到cuda上进行计算 classifier = classifier.eval() # 模型设置为评估模式 pred, _, _ = classifier(points) # print(pred.shape) # torch.Size([32, 16]) 输入有32个点云模型, 输出就有32*16大小的矩阵, 32行代表32个点云模型, 每行16个值代表可能属于16个类别的分数 loss = F.nll_loss(pred, target) pred_choice = pred.data.max(1)[1] # 按行寻找最大值的位置 correct = pred_choice.eq(target.data).cpu().sum().item() print('[%d: %d/%d] %s loss: %f accuracy: %f' % (epoch, i, num_batch, blue('test'), loss.item(), correct/float(opt.batchSize))) # print('n = ', n) torch.save(classifier.state_dict(), '%s/cls_model_%d.pth' % (opt.outf, epoch)) # 每个epoch都保存一个模型 total_correct = 0 total_testset = 0 for i,data in tqdm(enumerate(testdataloader, 0)): # enumerate在字典上是枚举、列举的意思, enumerate参数为可遍历/可迭代的对象(如列表、字符串), 后面的0是索引从0开始 points, target = data target = target[:, 0]