Upload 6 files

.gitattributes

@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+assets/bus.jpg filter=lfs diff=lfs merge=lfs -text

LICENSE

@@ -0,0 +1,28 @@
+BSD 3-Clause License
+
+Copyright (c) 2026, AXERA
+
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are met:
+
+1. Redistributions of source code must retain the above copyright notice, this
+   list of conditions and the following disclaimer.
+
+2. Redistributions in binary form must reproduce the above copyright notice,
+   this list of conditions and the following disclaimer in the documentation
+   and/or other materials provided with the distribution.
+
+3. Neither the name of the copyright holder nor the names of its
+   contributors may be used to endorse or promote products derived from
+   this software without specific prior written permission.
+
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
+DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE
+FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.

README.md

@@ -0,0 +1,18 @@
+# Deformable-Detr.axera
+Deformable-Detr DEMO on Axera NPU.
+
+### 1. 工程下载
+```  
+git clone https://github.com/AXERA-TECH/deformable-detr.axera.git
+```
+
+### 2. 模型转换
+```
+pulsar2 build --config  ./config/config.json
+```
+### 3. 板端运行
+```
+python inference.py --model detr.axmodel --img ./assets/bus.jpg --output out.jpg --thresh 0.6
+```
+### 4. 结果展示
+![result](./output/out.jpg)

config/config.json

@@ -0,0 +1,60 @@
+{
+    "input": "./onnx_model/deformable_detr_ax.onnx",
+    "output_dir": "./output",
+    "output_name": "detr.axmodel",
+    "target_hardware": "AX650",
+    "model_type": "ONNX",
+    "npuMode": "NPU3",
+    "onnx_opt": {
+        "disable_onnx_optimization": false,
+        "enable_onnxsim": false,
+        "model_check": false,
+        "disable_transformation_check": false
+    },
+    "quant": {
+        "precision_analysis": true,
+        "calibration_method": "MinMax",
+        "precision_analysis_method": "EndToEnd",
+        "precision_analysis_mode": "NPUBackend",
+        "input_configs": [
+            {
+                "tensor_name": "DEFAULT",
+                "calibration_dataset": "./data/images.tar.gz",
+                "calibration_mean": [123.675, 116.28, 103.53],
+                "calibration_std": [58.395, 57.12, 57.375]
+            }
+        ],
+        "layer_configs": [
+            {
+                "start_tensor_names": [
+                    "DEFAULT"
+                ],
+                "end_tensor_names": [
+                    "DEFAULT"
+                ],
+                "data_type": "U16"
+            }
+        ],
+        "conv_bias_data_type": "S32",
+        "transformer_opt_level": 0,
+        "enable_smooth_quant": false
+    },
+    "input_processors": [
+        {
+            "tensor_name": "input",
+            "src_format": "RGB",
+            "src_dtype": "FP32",
+            "src_layout": "NHWC",
+            "tensor_format": "RGB"
+        }
+    ],
+    "output_processors": [
+        {
+            "tensor_name": "DEFAULT",
+            "tensor_layout": "NCHW"
+        }
+    ],
+    "compiler": {
+        "check": 0
+    }
+}

data/images.tar.gz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:aad908e27b753be3daabc7db0cbeab657887b4a9b41d3232d87f4f98e4e4a3eb
+size 1094082

src/inference.py

@@ -0,0 +1,112 @@
+import argparse
+import numpy as np
+import sys
+import os
+
+try:
+    import axengine as ort
+    print("Running on AXera NPU (axengine)...")
+except ImportError:
+    import onnxruntime as ort
+    print("Running on CPU/GPU (onnxruntime)...")
+
+from PIL import Image, ImageDraw, ImageFont
+
+NORMALIZATION_ENABLED = False
+MEAN = np.array([123.675, 116.28, 103.53], dtype=np.float32)
+STD  = np.array([58.395, 57.12, 57.375], dtype=np.float32)
+
+CLASSES = [
+    "person", "bicycle", "car", "motorcycle", "airplane", "bus", "train", "truck", "boat", "traffic_light",
+    "fire_hydrant", "stop_sign", "parking_meter", "bench", "bird", "cat", "dog", "horse", "sheep", "cow",
+    "elephant", "bear", "zebra", "giraffe", "backpack", "umbrella", "handbag", "tie", "suitcase", "frisbee",
+    "skis", "snowboard", "sports_ball", "kite", "baseball_bat", "baseball_glove", "skateboard", "surfboard",
+    "tennis_racket", "bottle", "wine_glass", "cup", "fork", "knife", "spoon", "bowl", "banana", "apple",
+    "sandwich", "orange", "broccoli", "carrot", "hot_dog", "pizza", "donut", "cake", "chair", "couch",
+    "potted_plant", "bed", "dining_table", "toilet", "tv", "laptop", "mouse", "remote", "keyboard",
+    "cell_phone", "microwave", "oven", "toaster", "sink", "refrigerator", "book", "clock", "vase",
+    "scissors", "teddy_bear", "hair_drier", "toothbrush"
+]
+
+def preprocess_normalized(image_path, input_h, input_w, layout="NCHW"):
+    raw_image = Image.open(image_path).convert("RGB")
+    img_w, img_h = raw_image.size
+    
+    scale = min(input_w / img_w, input_h / img_h)
+    new_w, new_h = int(img_w * scale), int(img_h * scale)
+    resized_image = raw_image.resize((new_w, new_h), Image.BILINEAR)
+
+    canvas = Image.new("RGB", (input_w, input_h), (0, 0, 0)) 
+    canvas.paste(resized_image, (0, 0))
+    image_data = np.array(canvas, dtype=np.float32)
+
+    if NORMALIZATION_ENABLED:
+        image_data = (image_data - MEAN) / STD
+
+    if layout == "NCHW":
+        image_data = image_data.transpose(2, 0, 1)
+    
+    image_data = np.expand_dims(image_data, 0)
+    return image_data, raw_image, {"original_size": (img_w, img_h), "scale": scale}
+
+def main():
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--model", type=str, required=True)
+    parser.add_argument("--img", type=str, required=True)
+    parser.add_argument("--output", type=str, default="result.jpg")
+    parser.add_argument("--thresh", type=float, default=0.3)
+    opt = parser.parse_args()
+
+    session = ort.InferenceSession(opt.model)
+    input_meta = session.get_inputs()[0]
+    
+    if input_meta.shape[1] == 3:
+        layout, h, w = "NCHW", input_meta.shape[2], input_meta.shape[3]
+    else:
+        layout, h, w = "NHWC", input_meta.shape[1], input_meta.shape[2]
+
+    img_tensor, raw_img, meta = preprocess_normalized(opt.img, h, w, layout)
+    outputs = session.run(None, {input_meta.name: img_tensor})
+
+    dets = outputs[0][0]    
+    labels = outputs[1][0]  
+    scores = dets[:, 4]
+    keep = scores >= opt.thresh
+    
+    v_dets = dets[keep]
+    v_labels = labels[keep]
+
+    orig_w, orig_h = meta["original_size"]
+    scale = meta["scale"]
+
+    print(f"Detected {len(v_dets)} objects.")
+
+    if len(v_dets) > 0:
+        draw = ImageDraw.Draw(raw_img)
+        try:
+            font = ImageFont.truetype("DejaVuSans.ttf", 18)
+        except:
+            font = ImageFont.load_default()
+
+        for i in range(len(v_dets)):
+            box = v_dets[i, :4] / scale
+            score = v_dets[i, 4]
+            label_id = int(v_labels[i])
+            
+            x1, y1, x2, y2 = box
+            x1, x2 = np.clip([x1, x2], 0, orig_w)
+            y1, y2 = np.clip([y1, y2], 0, orig_h)
+
+            draw.rectangle([x1, y1, x2, y2], outline="lime", width=3)
+            
+            name = CLASSES[label_id] if label_id < len(CLASSES) else f"obj_{label_id}"
+            text = f"{name} {score:.2f}"
+            
+            draw.rectangle([x1, y1-20, x1+100, y1], fill="lime")
+            draw.text((x1+2, y1-20), text, fill="black", font=font)
+
+    raw_img.save(opt.output)
+    print(f"Result saved to {opt.output}")
+
+if __name__ == "__main__":
+    main()