From c66dc8b62b953c6953feac643a9536e6b2daa9cd Mon Sep 17 00:00:00 2001
From: thebears <thebears@orin.lan>
Date: Tue, 10 Jun 2025 18:06:09 -0400
Subject: [PATCH] py

---
 compare.py            |  53 +++++
 deepstream_obj_det.py | 471 ++++++++++++++++++++++++++++++++++++++++++
 try_decode.py         |  33 +++
 3 files changed, 557 insertions(+)
 create mode 100644 compare.py
 create mode 100755 deepstream_obj_det.py
 create mode 100644 try_decode.py

diff --git a/compare.py b/compare.py
new file mode 100644
index 0000000..3da58e7
--- /dev/null
+++ b/compare.py
@@ -0,0 +1,53 @@
+import numpy as np
+import json
+datum = np.load('dump.npz.npy')
+
+
+with open('dump.json','r') as rr:
+    js = json.load(rr)
+
+
+emb_dict = dict()
+for embed in js[1]:
+    fr = embed['frame_number']
+    vec = embed['vector']
+    emb_dict[fr] = np.asarray(vec)
+
+
+
+
+
+def cosine_sim(emb_in_1, emb_in_2):
+    emb_in_1 = emb_in_1.astype(np.float32)
+    emb_in_2 = emb_in_2.astype(np.float32)    
+    emb1_norm = np.linalg.norm(emb_in_1)
+    emb2_norm = np.linalg.norm(emb_in_2)
+    dot_prod = np.dot(emb_in_1, emb_in_2)
+    similarity = dot_prod/(emb1_norm * emb2_norm)
+
+    return np.round(emb1_norm,5),np.round( emb2_norm,5), np.round( dot_prod,5),np.round( similarity,5)
+
+
+arr_in_deepstream = np.asarray([y for _,y in emb_dict.items()])
+
+normed = np.divide(datum.T, np.linalg.norm(datum, axis=1)).T
+print('_________________________')
+
+for fr, emb in emb_dict.items():
+    emb1 = np.linalg.norm(emb)
+    emb2 = np.linalg.norm(datum[fr])
+#    print( cosine_sim(emb, datum[fr]))
+
+
+print('Deepstream and Actual norm')    
+print(np.max(np.dot(arr_in_deepstream, normed.T),axis=1))
+
+print('_________________________')
+for dat in datum:
+#    print(cosine_sim(dat, datum[0]))
+     pass
+
+
+#print(cosine_sim(datum[fr], datum[fr+1]))
+
+#print(cosine_sim(emb_dict[fr], emb_dict[fr+1]))
diff --git a/deepstream_obj_det.py b/deepstream_obj_det.py
new file mode 100755
index 0000000..6014d15
--- /dev/null
+++ b/deepstream_obj_det.py
@@ -0,0 +1,471 @@
+
+import sys
+sys.path.append('/opt/nvidia/deepstream/deepstream/sources/deepstream_python_apps/apps')
+import os
+import gi
+gi.require_version('Gst', '1.0')
+from gi.repository import GLib, Gst
+from common.platform_info import PlatformInfo
+from common.bus_call import bus_call
+import numpy as np
+import ctypes
+import pyds
+from functools import partial
+from CommonCode.settings import get_logger, LogColorize
+import argparse
+pfm = LogColorize.watch_and_fix_permissions
+
+logger = get_logger(__name__,'/var/log/ml_vision_logs/00_watch_and_fix_permissions', stdout=True, systemd=False)
+
+
+target_width = 1280
+target_height = 720
+os.environ.pop("DISPLAY",None)
+MUXER_BATCH_TIMEOUT_USEC = 1000000
+def embedder_results_probe(pad,info,u_data, list_add, frame_num = 0):
+    gst_buffer = info.get_buffer()
+    if not gst_buffer:
+        print("Unable to get GstBuffer ")
+        return
+    batch_meta = pyds.gst_buffer_get_nvds_batch_meta(hash(gst_buffer))
+    l_frame = batch_meta.frame_meta_list
+
+    while l_frame is not None:
+        try:
+            # Note that l_frame.data needs a cast to pyds.NvDsFrameMeta
+            # The casting also keeps ownership of the underlying memory
+            # in the C code, so the Python garbage collector will leave
+            # it alone.
+            frame_meta = pyds.NvDsFrameMeta.cast(l_frame.data)
+        except StopIteration:
+            break
+
+        frame_number=frame_meta.frame_num
+
+
+        l_user = frame_meta.frame_user_meta_list
+        while l_user is not None:
+            try:
+                # Note that l_user.data needs a cast to pyds.NvDsUserMeta
+                # The casting also keeps ownership of the underlying memory
+                # in the C code, so the Python garbage collector will leave
+                # it alone.
+                user_meta = pyds.NvDsUserMeta.cast(l_user.data)
+            except StopIteration:
+                break
+
+            if (
+                    user_meta.base_meta.meta_type
+                    != pyds.NvDsMetaType.NVDSINFER_TENSOR_OUTPUT_META
+            ):
+                continue
+
+            tensor_meta = pyds.NvDsInferTensorMeta.cast(user_meta.user_meta_data)
+
+            # Boxes in the tensor meta should be in network resolution which is
+            # found in tensor_meta.network_info. Use this info to scale boxes to
+            # the input frame resolution.
+            layers_info = []
+            if True:
+                for i in range(tensor_meta.num_output_layers):
+                    layer = pyds.get_nvds_LayerInfo(tensor_meta, i)
+                    if layer.layerName=='embedding':
+                        
+                        ptr = ctypes.cast(pyds.get_ptr(layer.buffer), ctypes.POINTER(ctypes.c_float))
+                        num_elements = layer.inferDims.numElements
+                        v = list(np.ctypeslib.as_array(ptr, shape=(num_elements,)))
+                        v = [float(x) for x in v]
+                        
+                        list_add.append({'frame_number':frame_number, 'vector':v})
+
+
+                            
+
+        
+                try:
+                    l_user = l_user.next
+                except StopIteration:
+                    break
+
+        try:
+            # indicate inference is performed on the frame
+            frame_meta.bInferDone = True
+            l_frame = l_frame.next
+        except StopIteration:
+            break            
+    
+
+    return Gst.PadProbeReturn.OK
+
+
+
+
+def detector_results_probe(pad,info,u_data, list_add, frame_num = 0):
+    frame_number=0
+    num_rects=0
+    got_fps = False
+
+    gst_buffer = info.get_buffer()
+    if not gst_buffer:
+        print("Unable to get GstBuffer ")
+        return
+
+
+    batch_meta = pyds.gst_buffer_get_nvds_batch_meta(hash(gst_buffer))
+    l_frame = batch_meta.frame_meta_list
+        
+    while l_frame is not None:
+        try:
+            frame_meta = pyds.NvDsFrameMeta.cast(l_frame.data)
+        except StopIteration:
+            break
+
+        frame_number=frame_meta.frame_num
+        l_obj=frame_meta.obj_meta_list
+        num_rects = frame_meta.num_obj_meta
+
+        l_user = frame_meta.frame_user_meta_list
+
+
+        
+        while l_obj is not None:
+            try: 
+                # Casting l_obj.data to pyds.NvDsObjectMeta
+                obj_meta=pyds.NvDsObjectMeta.cast(l_obj.data)
+            except StopIteration:
+                break
+
+            # param_extract = ['left','top','width','height']
+            # strc = ''
+            # for param in param_extract:
+            #     strc+=str(getattr(obj_meta.rect_params, param))
+            #     strc+=' '
+
+ #           target_width
+#            target_height
+            score = obj_meta.confidence
+            label = obj_meta.obj_label
+            left = obj_meta.rect_params.left
+            top = obj_meta.rect_params.top
+            width = obj_meta.rect_params.width
+            height = obj_meta.rect_params.height
+            frame_number = frame_number
+            class_id = obj_meta.class_id
+
+            d_add = {'score':score, 'label':label, 'left':left, 'top':top, 'width':width, 'height':height, 'frame_number':frame_number, 'class_id': class_id}
+            list_add.append(d_add)
+            print(frame_number, label, score)
+            if frame_number % 100 == 0:
+                str_pr = 'FRAME_PROGRESS: '+pfm(str(frame_number) + '/' + str(frame_num))
+                logger.info(str_pr)
+        
+
+            try: 
+                l_obj=l_obj.next
+            except StopIteration:
+                break
+
+        # Update frame rate through this probe
+        stream_index = "stream{0}".format(frame_meta.pad_index)
+
+        try:
+            l_frame=l_frame.next
+        except StopIteration:
+            break
+
+    return Gst.PadProbeReturn.OK
+
+
+
+def cb_newpad(decodebin, decoder_src_pad,data):
+    print("In cb_newpad\n")
+    caps=decoder_src_pad.get_current_caps()
+    if not caps:
+        caps = decoder_src_pad.query_caps()
+    gststruct=caps.get_structure(0)
+    gstname=gststruct.get_name()
+    source_bin=data
+    features=caps.get_features(0)
+
+    # Need to check if the pad created by the decodebin is for video and not
+    # audio.
+    print("gstname=",gstname)
+    if(gstname.find("video")!=-1):
+        # Link the decodebin pad only if decodebin has picked nvidia
+        # decoder plugin nvdec_*. We do this by checking if the pad caps contain
+        # NVMM memory features.
+        print("features=",features)
+        if features.contains("memory:NVMM"):
+            # Get the source bin ghost pad
+            bin_ghost_pad=source_bin.get_static_pad("src")
+            if not bin_ghost_pad.set_target(decoder_src_pad):
+                sys.stderr.write("Failed to link decoder src pad to source bin ghost pad\n")
+        else:
+            sys.stderr.write(" Error: Decodebin did not pick nvidia decoder plugin.\n")
+
+def decodebin_child_added(child_proxy,Object,name,user_data):
+    print("Decodebin child added:", name, "\n")
+    if(name.find("decodebin") != -1):
+        Object.connect("child-added",decodebin_child_added,user_data)
+
+    if "source" in name:
+        source_element = child_proxy.get_by_name("source")
+        if source_element.find_property('drop-on-latency') != None:
+            Object.set_property("drop-on-latency", True)
+
+
+def create_source_bin(uri):
+    print("Creating source bin")
+
+    # Create a source GstBin to abstract this bin's content from the rest of the
+    # pipeline
+    bin_name="source-bin-any-format"
+    print(bin_name)
+    nbin=Gst.Bin.new(bin_name)
+    if not nbin:
+        sys.stderr.write(" Unable to create source bin \n")
+
+    # Source element for reading from the uri.
+    # We will use decodebin and let it figure out the container format of the
+    # stream and the codec and plug the appropriate demux and decode plugins.
+    uri_decode_bin=Gst.ElementFactory.make("uridecodebin", "uri-decode-bin")
+    if not uri_decode_bin:
+        sys.stderr.write(" Unable to create uri decode bin \n")
+    # We set the input uri to the source element
+    uri_decode_bin.set_property("uri",uri)
+    # Connect to the "pad-added" signal of the decodebin which generates a
+    # callback once a new pad for raw data has beed created by the decodebin
+    uri_decode_bin.connect("pad-added",cb_newpad,nbin)
+    uri_decode_bin.connect("child-added",decodebin_child_added,nbin)
+
+    # We need to create a ghost pad for the source bin which will act as a proxy
+    # for the video decoder src pad. The ghost pad will not have a target right
+    # now. Once the decode bin creates the video decoder and generates the
+    # cb_newpad callback, we will set the ghost pad target to the video decoder
+    # src pad.
+    Gst.Bin.add(nbin,uri_decode_bin)
+    bin_pad=nbin.add_pad(Gst.GhostPad.new_no_target("src",Gst.PadDirection.SRC))
+    if not bin_pad:
+        sys.stderr.write(" Failed to add ghost pad in source bin \n")
+        return None
+    return nbin
+
+
+def run_inference(file_path):
+    os.environ.pop("DISPLAY",None)
+    if not file_path.startswith('file://'):
+        file_path = 'file://'+file_path
+
+    platform_info = PlatformInfo()
+    Gst.init(None)
+
+    pipeline = Gst.Pipeline()
+
+    streammux = Gst.ElementFactory.make("nvstreammux", "Stream-muxer")
+
+    nugget_detector = Gst.ElementFactory.make("nvinfer", "primary-inference")
+    nugget_embedder = Gst.ElementFactory.make("nvinfer", "secondary-inference")
+        
+    streammux.set_property('width', target_width)
+    streammux.set_property('height', target_height)
+    streammux.set_property('batched-push-timeout', MUXER_BATCH_TIMEOUT_USEC)
+    streammux.set_property('enable-padding',1)
+    streammux.set_property('batch-size', 4)
+
+
+#    nugget_detector.set_property('config-file-path', "/home/thebears/DeepStream-Yolo/detector.txt")
+    nugget_detector.set_property('config-file-path', "/home/thebears/DeepStream-Yolo/config_infer_primary_yoloV7.txt")
+    nugget_embedder.set_property('config-file-path', "/home/thebears/DeepStream-Yolo/embedder.txt")
+
+    fakesink1 = Gst.ElementFactory.make("fakesink","fakesink")
+    fakesink1.set_property('enable-last-sample', 0)
+    fakesink1.set_property('sync', 0)
+    pipeline.add(fakesink1)
+
+    fakesink2 = Gst.ElementFactory.make("fakesink","fakesink2")
+    fakesink2.set_property('enable-last-sample', 0)
+    fakesink2.set_property('sync', 0)
+    pipeline.add(fakesink2)
+    
+    pipeline.add(streammux)
+    pipeline.add(nugget_detector)
+    pipeline.add(nugget_embedder)
+
+
+
+#    uri_name =  'file:///home/thebears/railing.mp4'
+#    uri_name = 'file:///home/thebears/railing_00_20250213094806.mp4'
+    source_file=create_source_bin(file_path)
+
+    pipeline.add(source_file)
+    
+    stream_pad = streammux.request_pad_simple("sink_0")
+    source_pad = source_file.get_static_pad("src")
+    source_pad.link(stream_pad)
+
+    tee=Gst.ElementFactory.make("tee", "nvsink-tee")
+    pipeline.add(tee)
+
+    queue1=Gst.ElementFactory.make("queue", "nvtee-que1")
+    queue2=Gst.ElementFactory.make("queue", "nvtee-que2")
+    pipeline.add(queue1)
+    pipeline.add(queue2)
+    
+    streammux.link(tee)
+    
+
+    tee.link(queue1)
+    tee.link(queue2)
+
+     
+#    preprocess_detector = Gst.ElementFactory.make("nvdspreprocess","preprocess_detector")
+#    preprocess_detector.set_property('config-file', "/home/thebears/DeepStream-Yolo/detector_preprocess.txt")
+#    preprocess_detector.set_property('config-file',pre_file)
+
+    
+#    preprocess_embedder = Gst.ElementFactory.make("nvdspreprocess","preprocess_embedder")
+#    preprocess_embedder.set_property('config-file', "/home/thebears/DeepStream-Yolo/embedder_preprocess.txt")
+#    preprocess_embedder.set_property('config-file',pre_file)
+                                                  
+#    pipeline.add(preprocess_detector)
+#    pipeline.add(preprocess_embedder)
+
+#    queue1.link(preprocess_detector)
+#    preprocess_detector.link(nugget_detector)
+    
+#    queue2.link(preprocess_embedder)
+#    preprocess_embedder.link(nugget_embedder)
+    
+    
+    queue1.link(nugget_detector)
+    queue2.link(nugget_embedder)
+
+
+    cmd = f'/usr/bin/ffprobe -v error -select_streams v:0 -count_packets -show_entries stream=nb_read_packets -of csv=p=0 {file_path}'#/srv/ftp/railing/2025/02/28/railing_00_20250228115800.mp4
+    try:
+        frames = int(os.popen(cmd).read().strip())
+    except:
+        frames = 0
+
+    logger.info(f"TOTAL_FRAMES: {frames}")
+    
+    embedder_list = list()
+    
+
+    embedder_results = partial(embedder_results_probe, list_add=embedder_list, frame_num = frames)
+    nugget_embedder.get_static_pad("src").add_probe(Gst.PadProbeType.BUFFER, embedder_results, 0)
+
+
+    detector_list = list()
+    
+
+    detector_results = partial(detector_results_probe, list_add = detector_list, frame_num = frames)
+    nugget_detector.get_static_pad("src").add_probe(Gst.PadProbeType.BUFFER, detector_results, 0)
+
+
+
+    def get_pipeline_string(pipeline):
+        if not isinstance(pipeline, Gst.Pipeline):
+            return None
+
+        elements = []
+        iterator = pipeline.iterate_elements()
+        while True:
+            result, element = iterator.next()
+            if result != Gst.IteratorResult.OK:
+                break
+            elements.append(element.get_name())
+
+        return " ! ".join(elements)
+
+    nugget_detector.link(fakesink1)
+    nugget_embedder.link(fakesink2)
+
+
+    # create an event loop and feed gstreamer bus mesages to it
+    loop = GLib.MainLoop()
+    bus = pipeline.get_bus()
+    bus.add_signal_watch()
+    bus.connect ("message", bus_call, loop)
+    
+    # start play back and listen to events
+    print("Starting pipeline \n")
+    pipeline.set_state(Gst.State.PLAYING)
+    try:
+        loop.run()
+    except:
+        pass
+    # cleanup
+    pipeline.set_state(Gst.State.NULL)
+    return  detector_list, embedder_list
+
+def get_detailed_pipeline_string(pipeline):
+    """Generate a more detailed pipeline string with properties"""
+    if not isinstance(pipeline, Gst.Pipeline):
+        return None
+    
+    def get_element_string(element):
+        # Get element factory name
+        factory = element.get_factory()
+        if factory:
+            element_str = factory.get_name()
+        else:
+            element_str = element.get_name()
+        
+        # Add properties
+        props = []
+        for prop in element.list_properties():
+            # Skip some properties that are typically not set in command line
+            if prop.name in ('name', 'parent'):
+                continue
+            
+            try:
+                val = element.get_property(prop.name)
+                if val is not None and val != prop.default_value:
+                    # Format value appropriately based on type
+                    if isinstance(val, str):
+                        props.append(f"{prop.name}=\"{val}\"")
+                    elif isinstance(val, bool):
+                        props.append(f"{prop.name}={str(val).lower()}")
+                    else:
+                        props.append(f"{prop.name}={val}")
+            except:
+                # Skip properties that can't be read
+                pass
+                
+        if props:
+            element_str += " " + " ".join(props)
+            
+        return element_str
+
+    result = []
+    
+    # Simple approach - just gets top-level elements
+    iterator = pipeline.iterate_elements()
+    while True:
+        ret, element = iterator.next()
+        if ret != Gst.IteratorResult.OK:
+            break
+        result.append(get_element_string(element))
+    
+    return " ! ".join(result)
+
+
+
+    
+if __name__ == '__main__':
+    cpath  = sys.argv[1]
+    if cpath.endswith('-i'):
+        cpath = '/home/thebears/local/source/short.mp4'
+        
+    if not cpath.startswith('file'):
+        cpath = os.path.abspath(cpath)
+
+
+    out = run_inference(cpath)
+    
+    import json
+    with open('dump.json','w') as ff:
+        json.dump([out[0],out[1]],ff)
+    sys.exit()
+
diff --git a/try_decode.py b/try_decode.py
new file mode 100644
index 0000000..94ac97c
--- /dev/null
+++ b/try_decode.py
@@ -0,0 +1,33 @@
+
+import cv2
+
+#cap = cv2.VideoCapture("rtspsrc location=rtsp://wowzaec2demo.streamlock.net/vod/mp4:BigBuckBunny_115k.mov ! application/x-rtp, media=video ! rtph264depay ! h264parse ! nvv4l2decoder ! nvvidconv ! video/x-raw, format=BGRx ! videoconvert ! video/x-raw, format=BGR ! appsink", cv2.CAP_GSTREAMER)
+
+
+
+#cmd = 'filesrc location=/home/thebears/local/source/reduced.mp4 ! qtdemux name=demux demux.video_0 ! queue ! h264parse ! nvv4l2decoder ! nvvidconv ! appsink'
+
+cmd = 'filesrc location=/home/thebears/local/source/full.mp4 ! qtdemux name=demux demux.video_0 ! queue ! h265parse ! nvv4l2decoder ! nvvidconv ! appsink sync=false'
+#cmd = 'filesrc location=/home/thebears/local/source/full.mp4 ! qtdemux name=demux demux.video_0 ! h265parse ! avdec_h265 ! videoconvert ! appsink'
+
+
+#cmd = 'videotestsrc ! autovideosink'
+
+cap = cv2.VideoCapture(cmd, cv2.CAP_GSTREAMER)
+
+
+import time
+st = time.time()
+fr = 0
+while True:
+    good, frf = cap.read()
+    fr+=1
+    print(good, fr)
+    if not good:
+        break
+
+
+et = time.time()
+
+print(et-st, fr/(st-et))
+