Fixed build engines mode for pipeline start (#869)

insight-platform · Sep 30, 2024 · 0956cbc · 0956cbc
1 parent 7fe2b04
commit 0956cbc
Show file tree

Hide file tree

Showing 27 changed files with 96 additions and 33 deletions.
diff --git a/samples/age_gender_recognition/README.md b/samples/age_gender_recognition/README.md
@@ -10,8 +10,8 @@ Preview:
 
 Tested on platforms:
 
-- Xavier NX, Xavier AGX;
-- Nvidia Turing, Ampere.
+- Nvidia Turing
+- Nvidia Jetson Orin family
 
 Demonstrated operational modes:
 

diff --git a/samples/area_object_counting/README.md b/samples/area_object_counting/README.md
@@ -8,8 +8,8 @@ Preview:
 
 Tested on platforms:
 
-- Xavier NX, Xavier AGX;
-- Nvidia Turing, Ampere.
+- Nvidia Turing
+- Nvidia Jetson Orin family
 
 Demonstrated operational modes:
 

diff --git a/samples/auxiliary_streams/README.md b/samples/auxiliary_streams/README.md
@@ -2,6 +2,11 @@
 
 A pipeline demonstrating the use of Auxiliary Streams in Savant. The pipeline contains element, [Multiple Resolutions](multiple_resolutions.py). It scales the frame to multiple resolution and sends the frames to the auxiliary streams.
 
+Tested on platforms:
+
+- Nvidia Turing
+- Nvidia Jetson Orin family
+
 ## Prerequisites
 
 ```bash

diff --git a/samples/buffer_adapter/README.md b/samples/buffer_adapter/README.md
@@ -4,6 +4,11 @@ A pipeline demonstrates how Buffer Adapter works in Savant. In the demo video fr
 
 The buffer adapter metrics are stored in Prometheus and displayed on a Grafana dashboard.
 
+Tested on platforms:
+
+- Nvidia Turing
+- Nvidia Jetson Orin family
+
 ## Prerequisites
 
 ```bash

diff --git a/samples/conditional_video_processing/README.md b/samples/conditional_video_processing/README.md
@@ -6,6 +6,11 @@ Preview:
 
 ![](assets/conditional-video-processing.webp)
 
+Tested on platforms:
+
+- Nvidia Turing
+- Nvidia Jetson Orin family
+
 ## Prerequisites
 
 ```bash

diff --git a/samples/face_reid/README.md b/samples/face_reid/README.md
@@ -10,6 +10,11 @@ Preview:
 
 The sample is split into two parts: Index Builder and Demo modules.
 
+Tested on platforms:
+
+- Nvidia Turing
+- Nvidia Jetson Orin family
+
 ## Index Builder
 
 Index builder module loads images from [gallery](./assets/gallery), detects faces and facial landmarks, performs face preprocessing and facial recognition model inference. The resulting feature vectors are added into [hnswlib](https://github.com/nmslib/hnswlib) index, and the index (along with cropped face images from gallery) is saved on disk in the `index_files` directory.

diff --git a/samples/fisheye_line_crossing/README.md b/samples/fisheye_line_crossing/README.md
@@ -14,8 +14,8 @@ Preview:
 
 Tested on platforms:
 
-- Xavier NX, Xavier AGX;
-- Nvidia Turing, Ampere.
+- Nvidia Turing
+- Nvidia Jetson Orin family
 
 Demonstrated operational modes:
 

diff --git a/samples/intersection_traffic_meter/README.md b/samples/intersection_traffic_meter/README.md
@@ -10,8 +10,8 @@ Preview:
 
 Tested on platforms:
 
-- Xavier NX, Xavier AGX;
-- Nvidia Turing, Ampere.
+- Nvidia Turing
+- Nvidia Jetson Orin family
 
 Demonstrated operational modes:
 

diff --git a/samples/kafka_redis_adapter/README.md b/samples/kafka_redis_adapter/README.md
@@ -4,6 +4,11 @@ A pipeline demonstrates how Kafka-Redis adapters works in Savant. In the demo vi
 
 ![kafka-redis-adapter-demo.png](assets/kafka-redis-adapter-demo.png)
 
+Tested on platforms:
+
+- Nvidia Turing
+- Nvidia Jetson Orin family
+
 ## Prerequisites
 
 ```bash

diff --git a/samples/keypoint_detection/README.md b/samples/keypoint_detection/README.md
@@ -10,8 +10,8 @@ Preview:
 
 Tested on platforms:
 
-- Nvidia Jetson (Xavier NX, Xavier AGX, Orin family);
-- Nvidia Turing, Ampere.
+- Nvidia Turing
+- Nvidia Jetson Orin family
 
 Demonstrated adapters:
 

diff --git a/samples/kvs/README.md b/samples/kvs/README.md
@@ -2,6 +2,11 @@
 
 A pipeline demonstrates how to send frames to and receive from Kinesis Video Stream. Pipeline consists of two parts: exporter and importer. Exporter processes frames from a video file, sends metadata to MongoDB and sends frames to Kinesis Video Stream. Importer receives frames from Kinesis Video Stream, retrieves metadata from MongoDB and draw bboxes on frames.
 
+Tested on platforms:
+
+- Nvidia Turing
+- Nvidia Jetson Orin family
+
 ## Prerequisites
 
 ```bash

diff --git a/samples/license_plate_recognition/README.md b/samples/license_plate_recognition/README.md
@@ -8,8 +8,8 @@ Preview:
 
 Tested on platforms:
 
-- Xavier NX, Xavier AGX;
-- Nvidia Turing, Ampere.
+- Nvidia Turing
+- Nvidia Jetson Orin family
 
 Demonstrated adapters:
 

diff --git a/samples/mjpeg_usb_cam/README.md b/samples/mjpeg_usb_cam/README.md
@@ -4,6 +4,11 @@ A pipeline demonstrating how to capture MJPEG from a USB camera. MJPEG is a comm
 
 The resulting stream can be accessed via LL-HLS on `http://locahost:888/stream/video`
 
+Tested on platforms:
+
+- Nvidia Turing
+- Nvidia Jetson Orin family
+
 ## Hardware Acceleration Notes
 
 On X86, JPEG decoding and encoding is done in software (or hardware-assisted if dGPU and drivers support it). On Jetson, JPEG decoding and decoding is done in hardware with NVJPEG.

diff --git a/samples/multiple_gige/README.md b/samples/multiple_gige/README.md
@@ -4,6 +4,11 @@ A simple pipeline demonstrates how GigE Vision Source Adapter works in Savant. I
 
 The resulting streams can be accessed via LL-HLS on `http://locahost:888/stream/gige-raw` (raw-rgba frames) and `http://locahost:888/stream/gige-encoded` (HEVC-encoded frames).
 
+Tested on platforms:
+
+- Nvidia Turing
+- Nvidia Jetson Orin family
+
 Run the demo:
 
 ```bash

diff --git a/samples/multiple_rtsp/README.md b/samples/multiple_rtsp/README.md
@@ -4,6 +4,11 @@ A simple pipeline demonstrates how multiplexed processing works in Savant. In th
 
 The resulting streams can be accessed via LL-HLS on `http://locahost:888/stream/city-traffic` and `http://locahost:888/stream/town-centre`.
 
+Tested on platforms:
+
+- Nvidia Turing
+- Nvidia Jetson Orin family
+
 ## Prerequisites
 
 ```bash

diff --git a/samples/nvidia_car_classification/README.md b/samples/nvidia_car_classification/README.md
@@ -8,8 +8,8 @@ Preview:
 
 Tested on platforms:
 
-- Xavier NX, Xavier AGX;
-- Nvidia Turing, Ampere.
+- Nvidia Turing
+- Nvidia Jetson Orin family
 
 Demonstrated adapters:
 

diff --git a/samples/opencv_cuda_bg_remover_mog2/README.md b/samples/opencv_cuda_bg_remover_mog2/README.md
@@ -23,8 +23,8 @@ step-by-step [tutorial](https://blog.savant-ai.io/building-a-500-fps-accelerated
 
 Tested on platforms:
 
-- Xavier NX, Xavier AGX;
-- Nvidia Turing, Ampere.
+- Nvidia Turing
+- Nvidia Jetson Orin family
 
 Demonstrated operational modes:
 

diff --git a/samples/original_resolution_processing/README.md b/samples/original_resolution_processing/README.md
@@ -2,6 +2,11 @@
 
 A pipeline demonstrates processing of streams at the original resolution, i.e. without scaling to a single resolution. `parameters.frame` in  [module.yml](module.yml) is not specified. The sample sends two streams with resolutions 1280x720 and 1920x1080 to the module.
 
+Tested on platforms:
+
+- Nvidia Turing
+- Nvidia Jetson Orin family
+
 ## Prerequisites
 
 ```bash

diff --git a/samples/panoptic_driving_perception/README.md b/samples/panoptic_driving_perception/README.md
@@ -8,8 +8,8 @@ Preview:
 
 Tested on platforms:
 
-- Xavier NX, Xavier AGX;
-- Nvidia Turing, Ampere.
+- Nvidia Turing
+- Nvidia Jetson Orin family
 
 Demonstrated adapters:
 

diff --git a/samples/pass_through_processing/README.md b/samples/pass_through_processing/README.md
@@ -12,6 +12,11 @@ The modules performance is also stored in Prometheus and displayed on a Grafana
 
 Detector and tracker are running in pass-through mode (`codec: copy`). Draw-func encodes frames to H264.
 
+Tested on platforms:
+
+- Nvidia Turing
+- Nvidia Jetson Orin family
+
 ## Prerequisites
 
 ```bash

diff --git a/samples/peoplenet_detector/README.md b/samples/peoplenet_detector/README.md
@@ -28,8 +28,8 @@ step-by-step [tutorial](https://blog.savant-ai.io/meet-savant-a-new-high-perform
 
 Tested on platforms:
 
-- Xavier NX, Xavier AGX;
-- Nvidia Ampere.
+- Nvidia Turing
+- Nvidia Jetson Orin family
 
 Demonstrated operational modes:
 

diff --git a/samples/rtdetr/README.md b/samples/rtdetr/README.md
@@ -8,8 +8,8 @@ Weights used: `v0.1/rtdetr_r50vd_6x_coco_from_paddle.pth`  from the [RT-DETR rel
 
 Tested on platforms:
 
-- Nvidia Turing;
-- Nvidia Jetson Orin Nano.
+- Nvidia Turing
+- Nvidia Jetson Orin family
 
 Demonstrated operational modes:
 

diff --git a/samples/rtsp_cam_compatibility_test/README.md b/samples/rtsp_cam_compatibility_test/README.md
@@ -6,6 +6,11 @@ It uses NVDEC and NVENC internally and Savant protocol. Thus, if the pipeline wo
 
 The resulting video is broadcast in 640x360 resolution. You can access it at `http://<ip>:888/stream/test`.
 
+Tested on platforms:
+
+- Nvidia Turing
+- Nvidia Jetson Orin family
+
 ## Specifying the RTSP URL
 
 Edit `.env` file and set the `URI` variable to the RTSP URL of the camera.

diff --git a/samples/source_adapter_with_json_metadata/README.md b/samples/source_adapter_with_json_metadata/README.md
@@ -8,6 +8,11 @@ In the demo it is assumed that there is only one person in the picture and
 the IOU of the true box and from the Yolo detection model is calculated. 
 The IOU value is added as a tag to the frame metadata.
 
+Tested on platforms:
+
+- Nvidia Turing
+- Nvidia Jetson Orin family
+
 ## Prerequisites
 
 ```bash

diff --git a/samples/traffic_meter/README.md b/samples/traffic_meter/README.md
@@ -1,6 +1,5 @@
 # Traffic meter demo
 
-
 **NB**: The demo optionally uses **YOLOV8** model which takes up to **10-15 minutes** to compile to TensorRT engine. The first launch may take a decent time.
 
 The pipeline detects when people cross a user-configured line and the direction of the crossing. The crossing events are attached to individual tracks, counted for each source separately and the counters are displayed on the frame. The crossing events are also stored with Graphite and displayed on a Grafana dashboard.
@@ -17,8 +16,8 @@ Article on Medium: [Link](https://blog.savant-ai.io/efficient-city-traffic-meter
 
 Tested on platforms:
 
-- Xavier NX, Xavier AGX;
-- Nvidia Turing, Ampere.
+- Nvidia Turing
+- Nvidia Jetson Orin family
 
 Demonstrated operational modes:
 

diff --git a/samples/yolov8_seg/README.md b/samples/yolov8_seg/README.md
@@ -14,10 +14,10 @@ Preview:
 
 ![](assets/shuffle_dance.webp)
 
-
 Tested on platforms:
-- Xavier NX, Xavier AGX;
-- Nvidia Turing, Ampere.
+
+- Nvidia Turing
+- Nvidia Jetson Orin family
 
 ## Prerequisites
 

diff --git a/savant/deepstream/nvinfer/build_engine.py b/savant/deepstream/nvinfer/build_engine.py
@@ -35,16 +35,15 @@ def build_engine(element: ModelElement, rebuild: bool = True):
     pipeline: Gst.Pipeline = Gst.Pipeline()
     elements = [
         PipelineElement(
-            'videotestsrc',
-            properties={'num-buffers': model.batch_size},
+            'nvvideotestsrc',
+            properties={
+                'num-buffers': model.batch_size,
+            },
         ),
-        PipelineElement('nvvideoconvert'),
         PipelineElement(
             element='nvstreammux',
             name='muxer',
             properties={
-                'width': model.input.width if model.input.width else 1280,
-                'height': model.input.height if model.input.height else 720,
                 'batch-size': 1,
             },
         ),